Generative machine learning model for generating a blend of strains for food or beverage product fermentation

Abstract

The computer-implemented method (100) for providing at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, comprises: - a step of defining (135) at least one value representative of a post-fermentation food or beverage product performance target, - a step of generating (140) at least one blend digital identifier by a trained generative machine- learning model using as input each value representative of a post-fermentation food or beverage product performance target, - a step of providing (145) at least one of the generated at least one blend digital identifier.

Description

[0001] GENERATIVE MACHINE LEARNING MODEL FOR GENERATING A BLEND OF STRAINS FOR FOOD OR BEVERAGE PRODUCT FERMENTATION

[0002] Technical field of the invention

[0003] The present invention relates to a computer-implemented method for providing at least one blend digital identifier, representing a materialisable blend, associated with at least one strain digital identifier representative of an existing strain for food or beverage product fermentation, a computer- implemented method to train a generative machine learning model to provide at least one blend digital identifier, representing a materialisable blend, associated with at least one strain digital identifier representative of an existing strain for food or beverage product fermentation, and a corresponding computer program product, computer-readable storage medium and device.

[0004] The present invention is applicable to the domain food or beverage product design and more particularly when fermentation is implemented.

[0005] Background of the invention

[0006] Designing microbial strain blends for fermented food or beverage products is a complex task due to several reasons:

[0007] Strain Compatibility: Microbial strains must be compatible with each other to coexist and perform their fermentative roles effectively. Some strains may inhibit or enhance others in a process known as protocooperation, leading to an imbalance in the fermentation process, such as disclosed in Sieuwerts et al (2008) Unraveling Microbial Interactions in Food Fermentations: from

[0008] - Classical to Genomics Approaches. APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Vol. 74, No. 16: p. 4997-5007.

[0009] Specific Nutritional Requirements: Different strains have varied nutritional needs. A blend must provide a balanced environment where all strains can thrive without any one strain dominating due to preferential nutrient availability.

[0010] Optimal Growth Conditions: Each strain has its own optimal growth conditions, including temperature, pH, levels of oxygen and salinity. Designing a blend involves finding a compromise where all strains can grow effectively, but this can be challenging if their optimal conditions are significantly different.

[0011] Predictable Flavor Profiles: Fermentation significantly contributes to the flavor of products. Each microbial strain contributes differently to the flavor profile. Balancing these contributions to achieve a consistent and desirable flavor is a delicate and complex process. Safety Considerations: It's crucial to ensure that the microbial strains used are safe and do not produce harmful by-products. This involves rigorous testing and understanding of the metabolic pathways of each strain.

[0012] Stability and Consistency: The blend must remain stable over time and under various storage conditions. This stability ensures consistency in the final product across different batches and production cycles.

[0013] Scale-Up Challenges: Strains that work well in a laboratory setting may behave differently when scaled up to industrial levels. This scale-up can introduce unforeseen challenges in maintaining the balance and effectiveness of the blend.

[0014] Regulatory Compliance: There are strict regulations governing the use of microbial strains in food or beverage production. Ensuring that a strain blend complies with these regulations can be complex, especially when dealing with international markets with different standards.

[0015] Interaction with Food Matrix: The interaction between microbial strains and the specific food matrix (the physical, chemical, and biological environment of the food) can influence the fermentation process. This interaction needs to be carefully considered in blend design. Adaptation and Evolution: Microbial strains can adapt and evolve over time, which might alter their fermentative properties or interactions with other strains. Monitoring and adjusting the blend over time is necessary to maintain product quality.

[0016] Sensitivity to bacteriophages: several blends can have similar performance metrics however, these blends are designed to consist of strains harboring different phage resistance phenotypes, in order to prevent failure and product losses in industrial fermentations.

[0017] To overcome these challenges, extensive research and optimization are often required to identify compatible strains, understand their interactions, and fine-tune the fermentation conditions for optimal performance. There is no solution for generating blends responding to certain targets in timely fashion, without requiring intensive and skilled human labor.

[0018] Current approaches to overcome these challenges correspond to PCT patent application WO 2023 / 006778. Such approaches, however, is not suited for strictly digital generation of bacterial blends.

[0019] Current approaches to overcome these challenges correspond to US patent application US 2023 / 268034.

[0020] Current approaches to overcome these challenges correspond to US patent application US 2023 / 409975.

[0021] Current approaches to overcome these challenges correspond to scientific publication Leveraging knowledge engineering and machine learning for microbial bio-manufacturing”, by Oyetunde Tolutola et. Al, published in Biotechnology Advances, Elsevier Publishing, Barking, GB, Vol. 36, no. 4.

[0022] Such approaches are helpful for strain design, but offer no advantages for blends of strains.

[0023] Summary of the invention

[0024] The present invention is intended to remedy all or part of these disadvantages.

[0025] To this effect, according to a first aspect, the present invention aims at a computer-implemented method for providing at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, characterized in that it comprises: a step of defining at least one value representative of a post-fermentation food or beverage product performance target, a step of generating at least one blend digital identifier by a trained generative machine-learning model using as input each value representative of a post-fermentation food or beverage product performance target, a step of providing at least one generated blend digital identifier.

[0026] Thanks to these provisions, it is possible to obtain novel blend compositions that can be materialised and that will reach targets when materialised without requiring extensive laboratory exploration of strains and their interactions.

[0027] In particular embodiments, during the step of generating, a value representative of a quantity is generated for at least one strain represented by a strain digital identifier, said strain digital identifier being associated with at least one generated blend digital identifier.

[0028] Thanks to these provisions, it is possible to generate alternative blends defined by the quantities (either relative or absolute) of each strain. For example, two blends reaching a performance target can have the same additives in different amounts.

[0029] In particular embodiments, the computer-implemented method subject of the present invention comprises the steps of: providing an original set of exemplar blend digital identifiers, said set of exemplar blend digital identifiers being representative of materialised blends comprising at least two existing strains for food or beverage product fermentation, said blend digital identifiers being associated with at least one value representative of an empirically measured post-fermentation food or beverage product performance, to form a training set and training a generative machine-learning model using the training set, wherein the generative machine-learning model is trained to associate at least one value representative of an empirically measured post-fermentation food or beverage product performance with at least one strain digital identifier, wherein said trained generative machine-learning model is used during the step of generating. Thanks to these provisions, the generative machine-learning model is trained based on empirically measured, real-life, performances of different blends.

[0030] In particular embodiments the computer-implemented method subject of the invention comprises: a step of culturing a food or beverage product with a blend of at least two strains, a step of empirically measuring at least one value representative of a post-fermentation food or beverage product performance, a step of storing, in a database, said at least one value representative of an empirically measured post-fermentation food or beverage product performance in association with at least two strain digital identifiers representative of the at least two strains used during the step of culturing, wherein said database is used during the step of providing a training set.

[0031] Thanks to these provisions, the database constituting the training set is based on real-life samples and measured data from a tangible reality.

[0032] In particular embodiments, at least one value representative of a post-fermentation food or beverage product performance target is at least one of: a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, a potential of hydrogen after an elapsed predetermined fermentation time, a value representative of a texture of the post-fermentation food or beverage product, a value representative of a storage stability of the post-fermentation food or beverage product, a value representative of a microbial cell count at the end of fermentation and / or during shelf-life, and / or a value representative of a concentration of at least one chemical compound involved in the fermentation process.

[0033] The post-fermentation food or beverage product performance target may be similar in nature, but specific parameters may also be applicable to one type of either food or product performance. Such as the (absence of) turbidity, foaming, Brix measurement or refractive index.

[0034] Thanks to these provisions, the performance targets are adapted to real-life targets used in food or beverage product manufacturing. The above list is not limitative in any way and other postfermentation performance indicators may be used. In particular embodiments, at least one value representative of post-fermentation food or beverage product performance is a set of at least two different points representative of the evolution of the potential of hydrogen over time.

[0035] Thanks to these provisions, it is possible to find blends which match a specific time signature in terms of pH evolution.

[0036] In particular embodiments, the trained generative machine-learning model is a conditional variational auto-encoder (“CVAE”).

[0037] Thanks to these provisions, the generative machine-learning model is efficient and does not require and optimization loop.

[0038] In particular embodiments, during the step of generating: a quantity of at least one additional physical additive to be added to the blend is generated, said additive being represented by an additive digital identifier, and / or at least one culturing operational parameter of the fermentation process is generated. Thanks to these provisions, othertargets such as compatibility or necessity to add other strains or other additives (such as enzymes or vitamins) or the specification of the fermentation process can also be generated.

[0039] In particular embodiments the computer-implemented method subject of the invention further comprises a step of sending a digital command representative of an instruction of materialising at least one blend corresponding to at least one blend digital identifier provided.

[0040] Thanks to the provisions, the computer-implemented method can be linked to a device implementing the digital commands to materialise the blend.

[0041] In particular embodiments the computer-implemented method subject of the invention further comprises a step of materialising at least one blend corresponding to at least one blend digital identifier provided.

[0042] Thanks to these provisions, the blend is materialised.

[0043] According to a second aspect, the invention aims at a computer-implemented method to train a generative machine learning model to provide at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, characterized in that it comprises the steps of: providing an original set of exemplar blend digital identifiers, said set of exemplar blend digital identifiers being representative of materialised blends comprising at least two existing strains for food or beverage product fermentation, each existing strain for food or beverage product fermentation being represented by a strain digital identifier, said blend digital identifiers being associated with at least one value representative of an empirically measured postfermentation product performance, to form a training set, and training a generative machine-learning model using the training set, wherein the generative machine-learning model is trained to associate at least one value representative of an empirically measured post-fermentation food or beverage product performance with at least one strain digital identifier.

[0044] According to a third aspect, the invention aims at a computer program product comprising instructions which upon execution by a computer cause the computer to execute the method subject of the invention.

[0045] According to a fourth aspect, the invention aims at a computer-readable storage medium storing programming instructions which upon execution by a computer cause the computer to execute the method subject of the invention.

[0046] According to a fifth aspect, the invention aims at a device for providing at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, comprising: a means of defining at least one value representative of a post-fermentation food or beverage product performance target, a means of generating at least one blend digital identifier by a trained generative machine-learning model using as input each value representative of a post-fermentation food or beverage product performance target, a means of providing at least one of the generated at least one blend digital identifier. The second to fifth aspects of the present invention exhibit the same advantages as the related first aspect.

[0047] Detailed description of the drawings

[0048] Other advantages, purposes and particular characteristics of the invention shall be apparent from the following non-exhaustive description of at least one particular embodiment of the present invention, in relation to the drawings annexed hereto, in which:

[0049] [Figure 1] represents, schematically, a first succession of steps of a particular embodiment of the method subject of the present invention,

[0050] [Figure 2] represents, schematically, a computer system with which an embodiment of the method subject of the present invention can be implemented,

[0051] [Figure 3] represents, schematically, a conditional variational auto-encoder used in the method subject of the present invention, [Figure 4] represents, schematically, the training of the generative machine-learning model and use of said model in the method subject of the present invention,

[0052] [Figure 5] represents, schematically, the evolution of a potential of hydrogen level of a fermented food or beverage product over time,

[0053] [Figure 6] represents, schematically, a sampled curve of evolution of a potential of hydrogen level of a fermented food or beverage product over time,

[0054] [Figure 7] represents, schematically, the real derivative curve and a predicted curve of evolution of a potential of hydrogen level of a fermented food or beverage product over time,

[0055] [Figure 8] represents, schematically, the real curve and a predicted curve of evolution of a potential of hydrogen level of a fermented food or beverage product over time,

[0056] [Figure 9] represents, schematically, a histogram of sampling with distribution of real data,

[0057] [Figure 10] represents, schematically, a histogram of sampling with data generated by the trained generative machine-learning model subject of the invention, and

[0058] [Figure 11] represents, schematically, the correlation between the targeted and predicted data relating to the time to reach a potential of hydrogen level of 4.6 and the potential of hydrogen level after and elapsed time of 20 hours.

[0059] Detailed description of the invention

[0060] This description is not exhaustive, as each feature of one embodiment may be combined with any other feature of any other embodiment in an advantageous manner. Also, various inventive concepts may be embodied as one or more methods, of which an example has been provided. The acts performed as part of the method may be ordered in any suitable way. Accordingly, embodiments may be constructed in which acts are performed in an order different than illustrated, which may include performing some acts simultaneously, even though shown as sequential acts in illustrative embodiments.

[0061] The indefinite articles ‘a’ and ‘an’, as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean ‘at least one’.

[0062] The phrase ‘and / or’, as used herein in the specification and in the claims, should be understood to mean ‘either or both’ of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with ‘and / or’ should be construed in the same fashion, i.e. ‘one or more’ of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the ‘and / or’ clause whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, a reference to ‘A and / or B’, when used in conjunction with open-ended language such as ‘comprising’ can refer, in one embodiment, to A only (optionally including elements other than B); in another embodiment, to B only (optionally including elements other than A); in yet another embodiment, to both A and B (optionally including other elements); etc.

[0063] As used herein in the specification and in the claims, ‘or’ should be understood to have the same meaning as ‘and / or’ as defined above. For example, when separating items in a list, ‘or’ or ‘and / or’ shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as ‘only one of or ‘exactly one of, or, when used in the claims, ‘consisting of, will refer to the inclusion of exactly one element of a number or list of elements. In general, the term ‘or’ as used herein shall only be interpreted as indicating exclusive alternatives (i.e. ‘one or the other but not both’) when preceded by terms of exclusivity, such as ‘either,’ ‘one of,’ ‘only one of, or ‘exactly one of. ‘Consisting essentially of,’ when used in the claims, shall have its ordinary meaning as used in the field of patent law.

[0064] As used herein in the specification and in the claims, the phrase ‘at least one’, in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase ‘at least one’ refers, whether related or unrelated to those elements specifically identified. Thus, as a nonlimiting example, ‘at least one of A and B’ (or, equivalently, ‘at least one of A or B’, or, equivalently ‘at least one of A and / or B’) can refer, in one embodiment, to at least one, optionally including more than one, A, with no B present (and optionally including elements other than B); in another embodiment, to at least one, optionally including more than one, B, with no A present (and optionally including elements other than A); in yet another embodiment, to at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and optionally including other elements); etc.

[0065] In the claims, as well as in the specification above, all transitional phrases such as ‘comprising,’ ‘including,’ ‘carrying,’ ‘having,’ ‘containing,’ ‘involving,’ ‘holding,’ ‘composed of, and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases ‘consisting of and ‘consisting essentially of shall be closed or semi-closed transitional phrases, respectively.

[0066] It should be noted at this point that the figures are not to scale.

[0067] In the context of the present invention, “fermentation” refers to a metabolic process that converts carbohydrates, such as sugars and starches, into other chemical compounds, typically in the absence of oxygen. Microorganisms, such as bacteria, yeast, or fungi, carry out fermentation. This process is used in various industries, including food and beverage production, pharmaceuticals, and biofuel manufacturing. During fermentation, microorganisms break down complex organic compounds into simpler substances, releasing energy in the process. Many factors influence the fermentation process, from the strains used to the process parameters. Fermentation of products is a highly nonlinear process.

[0068] In the context of the present invention, a “fermented food or beverage product” refers to a type of edible or drinkable product that has undergone a controlled a fermentation process. In case of edible product-fermentation or beverage product-fermentation, the fermentation process leads to changes in its flavor, texture, nutritional profile, and shelf life. Examples of fermented edible or beverage products include:

[0069] Yogurt: Fermented dairy product produced by lactic acid bacteria, primarily Lactobacillus bulgaricus and Streptococcus thermophilus.

[0070] Sauerkraut: Fermented cabbage, often associated with European cuisine, produced by lactic acid bacteria.

[0071] Kimchi: A traditional Korean side dish made from fermented vegetables, typically Napa cabbage and radishes, flavored with various seasonings.

[0072] Sourdough Bread: Bread made with a naturally fermented dough, often using wild yeast and lactobacilli.

[0073] Cheese: Various types of cheese are produced through the fermentation of milk by specific strains of bacteria, molds, or yeast.

[0074] Tempeh: A traditional Indonesian product made from fermented soybeans, bound together into a compact cake.

[0075] Beer: One of the oldest and most widely consumed alcoholic beverages. It's typically made from malted barley, hops, yeast, and water, though various grains and flavorings can be used. Wine: Produced by fermenting crushed grapes or other fruits. The fermentation process converts sugars into alcohol, resulting in a wide range of flavors and styles.

[0076] Kombucha: A fermented tea beverage made by fermenting sweetened tea with a symbiotic culture of bacteria and yeast (SCOBY). It often has a slightly effervescent texture and can be flavored with fruits or herbs.

[0077] Kefir: A fermented milk drink made by adding kefir grains to milk. These grains contain a mixture of bacteria and yeast that ferment the lactose in the milk, creating a tangy, slightly carbonated beverage.

[0078] Sake: A traditional Japanese rice wine made by fermenting polished rice with koji mold and yeast. It has a unique flavor profile and can range from dry to sweet.

[0079] Cider: Made by fermenting apple juice, cider is popular in many regions, especially in areas with abundant apple orchards. It can be still or sparkling and may vary in sweetness. Mead: One of the oldest known alcoholic beverages, mead is made by fermenting honey with water and sometimes additional flavorings such as fruits, spices, or herbs.

[0080] Tepache: A traditional Mexican beverage made by fermenting pineapple rinds with brown sugar and spices. It has a sweet and tangy flavor with a hint of fermentation.

[0081] Jun: Similar to kombucha but made with green tea and honey instead of black tea and sugar. It often has a milder flavor compared to kombucha.

[0082] Lassi: A traditional South Asian yogurt-based drink that can be sweet or savory. It's often flavored with fruits, spices, or herbs and can be enjoyed plain or as a digestive aid.

[0083] In the context of the present invention, a “post-fermentation product performance target” refers to a goal to achieve that can be defined by parameters calculated or measured after the start of the fermentation process.

[0084] In the context of the present invention, a “strain” designates a strain of microorganisms that can be used for food or beverage product fermentation. Here are some common types of strains used in food product fermentation:

[0085] Lactic Acid Bacteria (“LAB”), such as:

[0086] Lactobacillus species: Common in dairy fermentations like yogurt, cheese, and sourdough such as Lactobacillus delbrueckii subsp. bulgaricus, L. helveticus, L. johnsonii, L. acidophilusn L. rhamnosus, L. casei, L. paracasei, L. plantarum, which can serve as acidifiers, as well as probiotics and / or bioprotective adjunct strains. Streptococcus species: Found in fermented dairy products and certain sausages. Leuconostoc mesenteroides.

[0087] Lactococcus

[0088] Yeasts:

[0089] Saccharomyces cerevisiae: Used in bread-making and alcoholic fermentations (beer, wine).

[0090] Candida milleri: Common in sourdough fermentation.

[0091] Wickerhamomyces anomalus: coffee fermentation

[0092] Acetic Acid Bacteria such as Acetobacter aceti: Used in the production of vinegar, converting ethanol into acetic acid.

[0093] Mold Strains:

[0094] Aspergillus oryzae: Used in the fermentation of soy sauce, miso, and sake. Penicillium roqueforti: Essential for blue cheese fermentation.

[0095] Koji Mold:

[0096] Aspergillus oryzae: Used in the fermentation of soy sauce, miso, and sake. It produces enzymes that break down starches and proteins. Propionibacteria:

[0097] Propionibacterium freudenreichii: Used in the production of Swiss cheese, contributing to the formation of characteristic holes and flavor.

[0098] Bacillus subtilis used in the fermentation of natto, a traditional Japanese soybean product. Wild Fermentation Strains naturally occurring strains that can be found on raw fruits and vegetables, contributing to spontaneous fermentation.

[0099] Probiotics such as Bifidobacterium

[0100] Here are some common types of strains used in beverage product fermentation: Saccharomyces cerevisiae: Commonly known as brewer's yeast, Saccharomyces cerevisiae is a key player in beer, wine, and cider production. It consumes sugars and produces alcohol and carbon dioxide as byproducts, crucial for fermentation.

[0101] Saccharomyces pastorianus: Used in lager beer production, this strain of yeast operates at lower temperatures compared to Saccharomyces cerevisiae. It contributes to the distinct flavors and characteristics of lager beers.

[0102] Brettanomyces: Often referred to as "Brett," this wild yeast strain can add unique flavors and aromas, including fruity, funky, and barnyard-like characteristics. It's commonly used in sour beer styles and certain wines.

[0103] Lactobacillus: This is a genus of bacteria commonly used in sour beer production, as well as in fermenting other beverages like kombucha. Lactobacillus produces lactic acid, contributing to sourness and acidity in the final product.

[0104] Pediococcus: Another type of bacteria often found in sour beer production. Like Lactobacillus, Pediococcus also produces lactic acid but typically operates at a slower rate. It contributes to the complexity of sour beer flavors.

[0105] Acetobacter: This bacteria is used in the production of vinegar and kombucha. It converts ethanol into acetic acid, which gives vinegar its sour taste.

[0106] Symbiotic Culture Of Bacteria and Yeast ("SCOBY”): It's a combination of various bacteria and yeast strains, including Acetobacter, Saccharomyces, and Lactobacillus, used to ferment sweet tea into kombucha.

[0107] Acetobacter xylinum: This specific strain of Acetobacter is used to produce the cellulose matrix in kombucha fermentation. It forms the physical structure of the SCOBY and aids in the fermentation process.

[0108] Saccharomyces bayanus: Another yeast species closely related to Saccharomyces cerevisiae, Saccharomyces bayanus is often used in winemaking, particularly for fermenting white wines and in some sparkling wine production. Non-Saccharomyces Yeasts: Besides Saccharomyces species, various non- Saccharomyces yeast species may also play a role in wine fermentation, contributing to the complexity of flavors and aromas. Examples include species from genera such as Hanseniaspora, Metschnikowia, and Pichia.

[0109] Lactic Acid Bacteria (Oenococcus oeni and Lactobacillus): These bacteria are responsible for conducting malolactic fermentation (MLF), a secondary fermentation process in winemaking. MLF converts harsh malic acid into softer lactic acid, contributing to the smoothness and complexity of many wines, particularly reds.

[0110] Wild Yeast Strains: In some traditional winemaking practices, wild or indigenous yeast strains naturally present on grape skins or in the winery environment may also play a role in fermentation. These wild yeasts can contribute unique flavors and characteristics to the finished wine.

[0111] Botrytis cinerea: Also known as noble rot, this fungus can affect grapes under specific climatic conditions, leading to the development of certain sweet wines such as Sauternes and Tokaji. Botrytis cinerea causes the grapes to shrivel, concentrating sugars and flavors in the berries.

[0112] Further common types of strains may be found in Ganzle (2022) The periodic table of fermented foods: limitations and opportunities. Applied Microbiology and Biotechnology [https: / / doi.org / 10.1007 / s00253-022-11909-y].

[0113] In the context of the present invention, the terms “strain digital identifier” refer to a digital bijective representation of an existing strain. Each strain digital identifier represents only one strain.

[0114] In the context of the present invention, the term “blend” refers to a blend of microorganism strains. In other words, it designates a combination of different microbial organisms designed for optimal performance in food or beverage product fermentation processes.

[0115] In the context of the present invention, the terms ‘blend digital identifier” refers to a digital bijective representation of a blend. Each blend digital identifier represents only one blend.

[0116] In the context of the present invention, the term ‘additive’ refers to any substance not normally consumed as a food or beverage in itself and not normally used as a characteristic ingredient of food, whether or not it has nutritive value. Added to food or beverages for technological purposes in its manufacture, processing, preparation, treatment, packaging, transport or storage, food or beverage additives become a component of the food or beverage, t. Such additives can be enzymes, hydrocolloids or vitamins, for example.

[0117] In the context of the present invention, the terms “culturing operation parameter” refer to any parameter which can be adjusted for the culturing operation. The specific parameters depend on the type of food or beverage product and the microorganisms involved. For example, the parameters that can be adjusted are at least one of:

[0118] Temperature: fermentation temperature is critical and varies based on the microorganisms used. Different strains have optimal temperature ranges for growth and activity.

[0119] Culturing Time: the duration of fermentation can be adjusted to achieve specific characteristics in the final product. Longer fermentation times often result in more complex flavors and textures.

[0120] Potential of hydrogen (“pH”) Level: pH plays a crucial role in microbial activity. Controlling pH levels can influence the dominance of certain microbial species and impact the flavor, texture, and safety of the fermented product.

[0121] Salt concentration: salt is commonly used in fermentation to control microbial activity and enhance the safety and flavor of the product. The concentration of salt can be adjusted to achieve the desired results.

[0122] Moisture Content: the moisture content of the fermenting substrate affects microbial growth and metabolism. Adjusting moisture levels can influence the texture and overall quality of the final product.

[0123] Oxygen Levels: some fermentations require anaerobic conditions (without oxygen), while others benefit from controlled exposure to oxygen. Adjusting the oxygen levels can impact the types of microorganisms that thrive during fermentation.

[0124] Inoculation Size: the amount of starter culture or inoculum added to the fermenting substrate can be adjusted to control the rate of fermentation and the population density of microorganisms.

[0125] Nutrient Availability: microorganisms require specific nutrients for growth. Modifying the nutrient composition of the substrate can impact microbial activity and the production of desirable compounds.

[0126] Adding of additives or ingredients: the dosage and timing can have an effect on the course of the food or beverage process, and the quality of the end product, such as the stability postfermentation, firmness, moisture content, and other characteristics of the final product.

[0127] Fermentation Vessel Type: the type of vessel used for fermentation (open, closed, traditional crock, modern fermenter) can influence the fermentation process and the characteristics of the final product.

[0128] Agitation: stirring or agitating the fermentation mixture can impact the distribution of microorganisms and nutrients, influencing the overall fermentation kinetics. Cultural Conditions: for some fermentations, such as those involving molds or specific bacteria, cultural conditions like humidity, light exposure, or the availability of specific gases might be adjusted.

[0129] A ‘quantity’ can be defined relatively to other components or in absolute units, moreover a quantity can refer to a weight or a concentration.

[0130] It should be noted that machine learning models, such as deep neural networks, have already allowed to model several aspects of the human cognition such as vision, hearing and language understanding. Such models have also been used on the generation of very realistic images and sounds / songs. However, models on the product fermentation have received very little attention, mainly because of the limited amount of data available.

[0131] The aim of the models presented herein is to generate new alternative blends of strains by considering certain targets, such as: a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, a potential of hydrogen after an elapsed fermentation time of twenty hours, a potential of hydrogen after an elapsed fermentation time of ten days, a value representative of a texture of the post-fermentation food or beverage product, and / or a value representative of a storage stability of the post-fermentation food or beverage product. at least two different points representative of the evolution of the potential of hydrogen over time.

[0132] In a general manner, the terms ‘digital representation identifier’ refer to any bijective digital representation of a physical item, such as a strain, a blend or an additive. Such a digital representation identifier may correspond to, for example, an entry in a database. A digital representation identifier may refer to a label representative of the name, chemical structure or internal reference of a strain, blend or additive, for example.

[0133] In the context of the present description, the term “materialized” is intended as existing outside of the digital environment of the present invention. ‘Materialised’ may mean, for example, readily found in nature or synthesized in a laboratory or chemical plant. In any event, a materialised blend digital identifier presents a tangible reality. The terms ‘to be compounded’ or ‘compounding’ refer to the act of materialisation of a blend digital identifier, whether via extraction and assembly of strains or via synthetization and assembly of strains.

[0134] As used herein, the terms “means of inputting” refer to, for example, a keyboard, mouse and / or touchscreen adapted to interact with a computing system in such a way to collect user input. In variants, the means of inputting are logical in nature, such as a network port of a computing system configured to receive an input command transmitted electronically. Such an input means may be associated to a GUI (Graphic User Interface) shown to a user or an API (Application programming interface). In other variants, the means of inputting may be a sensor configured to measure a specified physical parameter relevant for the intended use case. Examples of means of inputting are disclosed in regard to figure 2.

[0135] As used herein, the terms “computing system”, “computer”, or “computer system” designate any electronic calculation device, whether unitary or distributed, capable of receiving numerical inputs and providing numerical outputs by and to any sort of interface, digital and / or analog. Typically, a computing system designates either a computer executing a software having access to data storage or a client-server architecture wherein the data and / or calculation is performed at the server side while the client side acts as an interface. Examples of such computing systems are disclosed in regard to figure 2.

[0136] Figure 1 represents a computer-implemented method 100 for providing at least one blend digital identifier, representing a materialisable blend, associated with at least one strain digital identifier representative of an existing strain for product fermentation. Figure 1 represents different steps which can be grouped in four different phases: the first phase relates to the constitution of a database and comprises the following steps: a step of culturing 105 a food or beverage product with a blend of at least one strain, a step of empirically measuring 110 at least one value representative of a postfermentation food of beverage product performance, a step of storing 115, in a database, said at least one value representative of an empirically measured post-fermentation food or beverage product performance in association with at least one strain digital identifier representative of the at least one strain used during the step of culturing, said database is used during the step of providing a training set in the second phase, the second phase relates to the training of a generative machine-learning model and comprises the following steps: providing 125 an original set of exemplar blend digital identifiers, said set of exemplar blend digital identifiers being representative of materialised blends comprising at least one existing strain for food or beverageproduct fermentation, said blend digital identifiers being associated with at least one value representative of an empirically measured post-fermentation food or beverage product performance, to form a training set and training 130 a generative machine-learning model using the training set, wherein the generative machine-learning model is trained to associate at least one value representative of an empirically measured post-fermentation food or beverage product performance with at least one strain digital identifier, the third phase relates to the generation of at least one blend digital identifier comprising the following steps: a step of defining 135 at least one value representative of a post-fermentation food or beverage product performance target, a step of generating 140 at least one blend digital identifier by a trained generative machine-learning model using as input each value representative of a postfermentation food or beverage product performance target, a step of providing 145 at least one of the generated at least one blend digital identifier, the fourth phase relates to the materialising of at least one generated blend and comprises the following steps: a step of sending 150 a digital command representative of an instruction of materialising at least one blend corresponding to at least one blend digital identifier provided. a step of materialising 155 at least one blend corresponding to at least one blend digital identifier provided.

[0137] The first, second and fourth phases are optional embodiments of the method subject of the invention.

[0138] The first phase relates to the constitution of a database based on real world experimentation.

[0139] During the first phase, a blend of at least one strain is assembled and later added to a food or beverage product for culturing 105. Such a step is performed in a manner known to one skilled in the art. During the culturing process and / or after the culturing process, at least one measure 110 of a value representative of a post-fermentation food or beverage product performance is performed. Such measurements are performed in a manner known to one skilled in the art.

[0140] For example, at least one value measured can be: a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, preferably, the predetermined limit value is inferior to 5, more preferably, the predetermined limit value is equal to or less than 4.6, a potential of hydrogen after an elapsed fermentation time of twenty hours, a potential of hydrogen after an elapsed fermentation time of ten days, a value representative of a texture of the post-fermentation food or beverage product, and / or a value representative of a storage stability of the post-fermentation food or beverage product. For example, any measurement related to the pH can be performed using a pH meter. A texture of a product encompasses but is not limited to gel strength, viscosity, friction / lubrication, ropiness, extensional rheology, each of which can be measured empirically. For example, viscosity can be measured using a viscometer.

[0141] Preferably, at least one measurement is performed over time so as to define the evolution of a value over time. Even more preferably, at least one value representative of post-fermentation food or beverage product performance is at least two different points representative of the evolution of the potential of hydrogen over time. For example, during the step of empirically measuring 110, at least one value is measured at regular time intervals. Such time intervals can be every minute or every ten minutes, for example.

[0142] A curve 503 representing an empirically measured the evolution of the pH overtime is shown in figure 5. In figure 5, the abscissa 501 is a number of timesteps, each timestep being equal to one minute, the ordinate 602 is a pH level.

[0143] During the step of storing 115, each blend used during the step of culturing can correspond to a blend digital identifier. The database associates said blend digital identifier with at least one strain digital identifier of said blend and a value representative of post-fermentation food or beverage product performance. The database can also store one of the following elements: each strain combined in the blend, using a strain digital identifier, a value representative of a quantity of at least one strain of the blend, an information related to the storing of at least one strain of the blend and / or the blend, an information related to the experimental conditions such as a culturing operational parameter as defined here above and / or a quantity of at least one additional physical additive to add to the blend, said additive being represented by an additive digital identifier.

[0144] After the step of storing 115, the database is used during the step of providing 125 a training set of the second phase. The generative machine-learning model is therefore trained based on at least part of the elements constituting the database stated above.

[0145] Figure 1 also shows a computer-implemented method 120 to train a generative machine learning model to provide at least one blend digital identifier, representing a materialisable blend, associated with at least one strain digital identifier representative of an existing strain for food or beverage product fermentation, which comprises the steps of: providing 125 an original set of exemplar blend digital identifiers, said set of exemplar blend digital identifiers being representative of materialised blends comprising at least one existing strain for food or beverage product fermentation, each existing strain for food or beverage product fermentation being represented by a strain digital identifier, said blend digital identifiers being associated with at least one value representative of an empirically measured postfermentation food or beverage product performance, to form a training set, and training 130 a generative machine-learning model using the training set, wherein the generative machine-learning model is trained to associate at least one value representative of an empirically measured post-fermentation food or beverage product performance with at least one strain digital identifier.

[0146] The steps of the method 120 are described here below as being an optional embodiment of the method 100 subject of the present invention. However, such method 120 for training can be a method object of the present invention in itself, independently from the method 100 for providing at least one blend digital identifier.

[0147] The step of providing 125 is performed, for example, by a computer program executed by an electronic computation device, such as a microprocessor. During this step of providing 125, at least one blend digital identifier is provided to the generative machine-learning model to be trained. Such a step of providing 125 may correspond to the transfer in memory of the exemplar blend digital identifier from a digital storage location to another, the latter being dedicated to the training set of the generative machine-learning model. During such a step of providing 125, a larger sample of blend digital identifiers representative of distinct blend may be divided into a training set and into a validation set.

[0148] The step of training 130 may use any training method appropriated for training a generative machine learning model.

[0149] In preferred embodiments, the generative machine-learning model to be trained in a conditional variational auto-encoder (“CVAE”). A representation of a CVAE to be trained 305 is modeled in figure 3. Figure 3 also represents the trained generative machine-learning model 310 used during the step of generating.

[0150] An autoencoder is a specific type of neural network with typically the same number of neurons in the input and the output. The output is expected and enforced to be as close as possible to the input during the training process. The aim of that ‘copying machine’ is based on the interesting feature that there is a bottleneck in one of the layers of the neural network that serves the purpose of compressing the information represented in the input. Such bottleneck layer represents what is called the latent space and every neural network layer (also called hidden layer) before that bottleneck layer aims at compressing the information of the input in the most performative way with a much smaller number of neurons than the input, such that the most prominent features of the initial input are represented in the latent space. This is also called the ‘encoder’ part of the autoencoder. Every hidden layer after the bottleneck layer aims at decompressing the information of the latent space such that the output is as similar possible to the input as possible. Such part of the autoencoder is called the ‘decoder’. The way the output is enforced to be as similar to the output as possible is through what is called the ‘reconstruction loss’ that compares quantity by quantity of the input (typically represented by a vector) to the outputs and sums up the differences. The aim of the learning process is to get the lowest possible number for that sum through a process called backpropagation.

[0151] Autoencoders are used in different tasks, such as in denoising images, also called denoising autoencoders. Denoising autoencoders, after learning the sparse representations of the latent space, can be presented with noisy images and are able to denoise them through the latent space learned, as they have learned to distinguish relevant features from noise.

[0152] Another kind of autoencoder that is specifically interesting to the current invention is variational autoencoders. Variational autoencoders not only learn the sparse representations of the latent space but are also able to generate new outputs as well.

[0153] What the autoencoders are learning is the data distribution of the inputs and what the variational autoencoders do is to split that problem into two subproblems: learn the median vector and the standard deviation vector of the latent space distribution, imposing that the latent space distribution is Gaussian (otherwise, we cannot ensure that a median or standard deviation exists). Once the training process is finished, a new generated output can be provided by sampling randomly from the distribution learned in the latent space and feeding it to the decoder, generating a completely new output that comes from a continuous representation of the data distribution of which the model was trained off. Such a sample is provided by a sampler designed to generate randomized and unique values.

[0154] To train the variational autoencoder, one needs to consider a final trick that is called the reparameterization trick. The sampling operation that is fed to the decoder cannot be trained through backpropagation, so in order to properly train the variational autoencoder one considers a deterministic mean and standard deviation, and the sampling processing is inserted in a multiplication of a random vector that has a Gaussian distribution with the deterministic standard deviation. Such a random vector does not go through the backpropagation process, but only the median and standard deviation value. Therefore, the sampling process is split through a deterministic part that is learned and a stochastic part that is fixed (and therefore not learned). Once the training is done, one can sample a random vector from a normal distribution and feed it to the decoder.

[0155] Variational autoencoders can be improved thanks to the quantization of the latent space that allows to employ what is called autoregressive models. The idea of such autoregressive models is that the generation is conditioned on its own previous values, therefore the first value of the output vector is a random value, the second value of the output is conditioned on the first value generated, the third value depends on the first and second value, etc. Such autoregressive models play the role of the decoder after the training process is finished. An example architecture for a variational autoencoder training is a fully connected neural network comprising 1 to 5 hidden layers with each layer having a number of neurons equal to the previous layer’s output. In particular embodiments, it is possible to apply non-linearity (any activation function like relu, prelu, elu, selu, swish, mish, sigmoid, tanh, tanhexp,) to some or all hidden layers and optionally to the output layer.

[0156] The conditional VAE (CVAE), is a subtype of variational auto-encoders in which label information, such as performance targets, is inserted in the latent space to force a deterministic constrained representation of the learned data.

[0157] In figure 3, the variables of the CVAE to be trained 305 are:

[0158] “x” represents the blend digital identifier,

[0159] “w” the at least one value representative of a post-fermentation food or beverage product performance target, otherwise known as “target”,

[0160] “z” the latent space representation,

[0161] “p” is the mean,

[0162] “o” is the standard deviation,

[0163] “fe(w,x)” is the distribution of the encoder,

[0164] “ge(w,x)” is the distribution of the decoder, which is trained during the training, and

[0165] “N(0,l)” is the noise inserted between the encoder and the decoder, depending on the number of layers in the latent space.

[0166] As can be seen from figure 3, once the CVAE is trained 310, the decoder function is used to generate a least one blend digital identifier using defined targets.

[0167] When data stored in the database such as a curve 503 representing an empirically measured the evolution of the pH overtime is used for training the CVAE, the curve can be sampled 603, as represented in figure 6. In figure 6, the abscissa 601 is a number of timesteps, each timestep being equal to ten minutes, the ordinate 602 is a pH level, the curve is then dimensionally reduced via principal component analysis (“PCA”) such that the data is linearly transformed onto a new coordinate system such that the directions, also known as “principal components”, capturing the largest variation in the data can be easily identified. The curve 503 is sampled in a curve 603 that is a sequence of adjoining straight lines.

[0168] Figure 7 represents two curves 703 and 704, the curve 703 represents the curve 503 after mean removal from the curve 503, the curve 704 represents the curve 603 after mean removal from the curve 603 also known as “reconstructed curve”. In figure 7, the abscissa 701 is a number of timesteps, each timestep being equal to ten minutes, the ordinate 602 is a pH level. As can be seen, both curves are very similar showing that the sampled curve is representative of the empirically measured evolution of pH overtime. Thus, the sampling can be used for training without overburdening the database.

[0169] By using empirically measured the evolution of the pH overtime is used for training the CVAE sampled as explained above, it is possible to train the CVAE subject of the invention to predict curve values.

[0170] The trained generative machine-learning model represented in figure 3 is then used during the third phase.

[0171] Figure 4 represents the training 400 of the generative machine-learning model. During the training phase 401 , i.e., phase 120 represented in figure 1 , the model is fed data from a database of blend digital identifiers 402 to explicitly define a target 403 to blend digital identifier dependency. Such a target can be modeled by key performance indicators otherwise defined in the present description as values representative of a post-fermentation food or beverage product performance. It should be noted that one target can be achieved by multiple different blend digital identifiers.

[0172] A conditional distribution representing the likelihood of the blend digital identifier with the target is modeled and sampled from.

[0173] In preferred embodiments, in order to optimize 404 the distribution, the generated blend digital identifiers are used as a starting point for the training of the generative machine-learning model 400. Such a training makes it possible to control and / or enforce diversity in the generated blend digital identifiers.

[0174] Once the model is trained, at least one target 405 can be defined, the trained generative machine-learning model 400 generating at least one blend digital identifier.

[0175] Figure 9 represents a histogram 900 of sampling with distribution of curves of pH evolution over time sampled as explained with regards to figure 6 and figure 10 represents a histogram 1000 of sampling with similar data generated by the trained generative machine-learning model represented in figure 3. As can be seen, the training is efficient as the training and test error are roughly the same as those obtained with the data generated by the trained generative machine-learning model 310. The histograms 900 and 1000 show that curves can be accurately predicted by the trained generative machine-learning model 310. As such, values can be extracted from the generated curves such as: a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, a potential of hydrogen after an elapsed fermentation time of twenty hours, a potential of hydrogen after an elapsed fermentation time of ten days,

[0176] The third phase of the method represented in figure 1 , relates to the generation and provision of at least one blend digital identifier by the trained generative machine-learning model 400 to reach a target. Preferably, at least one value representative of a post-fermentation food or beverage product performance target is at least one of: a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, a potential of hydrogen after an elapsed predetermined fermentation time, a value representative of a texture of the post-fermentation food or beverage product, a value representative of a storage of the post-fermentation food or beverage product, a value representative of a cell count at the end of fermentation and / or during shelf-life, and / or a value representative of a concentration of at least one chemical compound involved in the fermentation process. at least two different points representative of the evolution of the potential of hydrogen over time.

[0177] It should be noted that a chemical compound involved in the fermentation process defines a chemical compound already present in the food or in the blend before culturing, or a chemical compound issued from a chemical reaction during the fermentation process.

[0178] The step of generating 140 is performed, for example, by a computer program executed upon an electronic computing unit or computing system, such as a computer, for example. During this step, the trained generative machine learning model is executed to provide, as an output, a blend digital identifier representative of a materialisable blend of strain for food or beverage product fermentation fitting predetermined values for the target (such as the values mentioned above).

[0179] In preferred embodiments, during the step of generating 140, a value representative of a quantity is generated for at least one strain represented by a strain digital identifier, said strain digital identifier being associated with at least one generated blend digital identifier. In such embodiments, the composition of the blend represented by the blend digital identifier is partially, preferably entirely generated.

[0180] In preferred embodiments, during the step of generating 140: a quantity of at least one additional physical additive to be added to the blend is generated, said additive being represented by an additive digital identifier, and / or at least one culturing operational parameter of the fermentation process is generated as defined above.

[0181] The additive and culturing operation parameter of the fermentation process are defined above. Figure 8 represents a curve 803 representing an empirically measured the evolution of the pH over time of a predefined blend as shown in figure 5 and a curve of a predicted evolution of pH for said predefined strain. In figure 8, the abscissa 801 is a number of timesteps, each timestep being equal to one minute, the ordinate 802 is a pH level. As can be seen, the predicted curve represents accurately the evolution of the pH overtime of said predefined strain. From the predicted curve target values can be extracted, such as : a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, a potential of hydrogen after an elapsed fermentation time of twenty hours, a potential of hydrogen after an elapsed fermentation time of ten days,

[0182] Figure 11 shows: in graph 1105, the correlation between the targeted and predicted values of time to reach a pH level of 4.6 and in graph 1110, the correlation between the targeted and predicted values of the pH level after an elapsed fermentation time of twenty hours.

[0183] Figure 11 shows the accuracy of the trained generative machine-learning model used in the method of the present invention as well as the corresponding training method.

[0184] The step of providing 145 can be performed analogically to the related steps of providing 125 of the method 100.

[0185] The fourth optional phase can comprise a step of selecting (not represented) at least one blend digital identifier provided. During the step of sending 150, a digital command representative of an instruction of materialising at least one blend corresponding to at least one blend digital identifier provided is generated and send to a device capable of materialising the at least one blend corresponding to at least one blend digital identifier provided.

[0186] The step of materialising may be performed manually, via a user interface for example, or automatically, via the device capable of materialising the at least one blend corresponding to at least one blend digital identifier provided based on the command sent.

[0187] Figure 2 represents a block diagram that illustrates an example computer system 200 with which an embodiment of the present invention may be implemented. In the example of figure 8, a computer system 205 and instructions for implementing the disclosed technologies in hardware, software, or a combination of hardware and software, are represented schematically, for example as boxes and circles, at the same level of detail that is commonly used by persons of ordinary skill in the art to which this disclosure pertains for communicating about computer architecture and computer systems implementations.

[0188] The computer system 205 includes an input / output (IO) subsystem 220 which may include a bus and / or other communication mechanism(s) for communicating information and / or instructions between the components of the computer system 205 over electronic signal paths. The I / O subsystem 220 may include an I / O controller, a memory controller and at least one I / O port. The electronic signal paths are represented schematically in the drawings, for example as lines, unidirectional arrows, or bidirectional arrows. At least one hardware processor 210 is coupled to the I / O subsystem 220 for processing information and instructions. Hardware processor 210 may include, for example, a general-purpose microprocessor or microcontroller and / or a special-purpose microprocessor such as an embedded system or a graphics processing unit (GPU) or a digital signal processor or ARM processor. Processor 210 may comprise an integrated arithmetic logic unit (ALU) or may be coupled to a separate ALU.

[0189] Computer system 205 includes one or more units of memory 225, such as a main memory, which is coupled to I / O subsystem 220 for electronically digitally storing data and instructions to be executed by processor 210. Memory 225 may include volatile memory such as various forms of random-access memory (RAM) or other dynamic storage device. Memory 225 also may be used for storing temporary variables or other intermediate information during execution of instructions to be executed by processor 210. Such instructions, when stored in non-transitory computer-readable storage media accessible to processor 210, can render computer system 205 into a special-purpose machine that is customized to perform the operations specified in the instructions.

[0190] Computer system 205 further includes non-volatile memory such as read only memory (ROM) 230 or other static storage device coupled to the I / O subsystem 220 for storing information and instructions for processor 210. The ROM 230 may include various forms of programmable ROM (PROM) such as erasable PROM (EPROM) or electrically erasable PROM (EEPROM). A unit of persistent storage 215 may include various forms of non-volatile RAM (NVRAM), such as FLASH memory, or solid-state storage, magnetic disk, or optical disk such as CD-ROM or DVD-ROM and may be coupled to I / O subsystem 220 for storing information and instructions. Storage 215 is an example of a non-transitory computer-readable medium that may be used to store instructions and data which when executed by the processor 210 cause performing computer-implemented methods to execute the techniques herein.

[0191] The instructions in memory 225, ROM 230 or storage 215 may comprise one or more sets of instructions that are organized as modules, methods, objects, functions, routines, or calls. The instructions may be organized as one or more computer programs, operating system services, or application programs including mobile apps. The instructions may comprise an operating system and / or system software; one or more libraries to support multimedia, programming or other functions; data protocol instructions or stacks to implement TCP / IP, HTTP or other communication protocols; file format processing instructions to parse or render files coded using HTML, XML, JPEG, MPEG or PNG; user interface instructions to render or interpret commands for a graphical user interface (GUI), command-line interface or text user interface; application software such as an office suite, Internet access applications, design and manufacturing applications, graphics applications, audio applications, software engineering applications, educational applications, games or miscellaneous applications. The instructions may implement a web server, web application server or web client. The instructions may be organized as a presentation layer, application layer and data storage layer such as a relational database system using structured query language (SQL) or no SQL, an object store, a graph database, a flat file system or other data storage.

[0192] Computer system 205 may be coupled via I / O subsystem 220 to at least one output device 235. In one embodiment, output device 235 is a digital computer display or Human Machine Interface. Examples of a display that may be used in various embodiments include a touchscreen display or a light-emitting diode (LED) display or a liquid crystal display (LCD) or an e-paper display. Computer system 205 may include other type(s) of output devices 235, alternatively or in addition to a display device. Examples of other output devices 235 include printers, ticket printers, plotters, projectors, sound cards or video cards, speakers, buzzers or piezoelectric devices or other audible devices, lamps or LED or LCD indicators, haptic devices, actuators, or servos.

[0193] At least one input device 240 is coupled to I / O subsystem 220 for communicating signals, data, command selections or gestures to processor 210. Examples of input devices 240 include touchscreens, microphones, still and video digital cameras, alphanumeric and other keys, keypads, keyboards, graphics tablets, image scanners, joysticks, clocks, switches, buttons, dials, slides.

[0194] Another type of input device is a control device 245, which may perform cursor control or other automated control functions such as navigation in a graphical interface on a display screen, alternatively or in addition to input functions. Control device 245 may be a touchpad, a mouse, a trackball, or cursor direction keys for communicating direction information and command selections to processor 210 and for controlling cursor movement on display 235. The input device may have at least two degrees of freedom in two axes, a first axis (e.g., x) and a second axis (e.g., y), that allows the device to specify positions in a plane. Anothertype of input device is a wired, wireless, or optical control device such as a joystick, wand, console, steering wheel, pedal, gearshift mechanism or other type of control device. An input device 240 may include a combination of multiple different input devices, such as a video camera and a depth sensor.

[0195] In another embodiment, computer system 205 may comprise an Internet of things (loT) device in which one or more of the output device 235, input device 240, and control device 245 are omitted. Or, in such an embodiment, the input device 240 may comprise one or more cameras, motion detectors, thermometers, microphones, seismic detectors, other sensors or detectors, measurement devices or encoders and the output device 235 may comprise a special-purpose display such as a single-line LED or LCD display, one or more indicators, a display panel, a meter, a valve, a solenoid, an actuator or a servo.

[0196] Computer system 205 may implement the techniques described herein using customized hardwired logic, at least one ASIC or FPGA, firmware and / or program instructions or logic which when loaded and used or executed in combination with the computer system causes or programs the computer system to operate as a special-purpose machine. According to one embodiment, the techniques herein are performed by computer system 205 in response to processor 210 executing at least one sequence of at least one instruction contained in main memory 225. Such instructions may be read into main memory 225 from another storage medium, such as storage 215. Execution of the sequences of instructions contained in main memory 225 causes processor 210 to perform the process steps described herein. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions.

[0197] The term “storage media” as used herein refers to any non-transitory media that store data and / or instructions that cause a machine to operate in a specific fashion. Such storage media may comprise non-volatile media and / or volatile media. Non-volatile media includes, for example, optical or magnetic disks, such as storage 215. Volatile media includes dynamic memory, such as memory 225. Common forms of storage media include, for example, a hard disk, solid state drive, flash drive, magnetic data storage medium, any optical or physical data storage medium, memory chip, or the like.

[0198] Storage media is distinct from but may be used in conjunction with transmission media. Transmission media participates in transferring information between storage media. For example, transmission media includes coaxial cables, copper wire and fiber optics, including the wires that comprise a bus of I / O subsystem 220. Transmission media can also take the form of acoustic or light waves, such as those generated during radio-wave and infra-red data communications.

[0199] Various forms of media may be involved in carrying at least one sequence of at least one instruction to processor 210 for execution. For example, the instructions may initially be carried on a magnetic disk or solid-state drive of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a communication link such as a fiber optic or coaxial cable or telephone line using a modem. A modem or router local to computer system 205 can receive the data on the communication link and convert the data to a format that can be read by computer system 205. For instance, a receiver such as a radio frequency antenna or an infrared detector can receive the data carried in a wireless or optical signal and appropriate circuitry can provide the data to I / O subsystem 220 such as place the data on a bus. I / O subsystem 220 carries the data to memory 225, from which processor 210 retrieves and executes the instructions. The instructions received by memory 225 may optionally be stored on storage 215 either before or after execution by processor 210.

[0200] Computer system 205 also includes a communication interface 260 coupled to bus 220. Communication interface 260 provides a two-way data communication coupling to network link(s) 265 that are directly or indirectly connected to at least one communication network, such as a network 270 or a public or private cloud on the Internet. For example, communication interface 260 may be an Ethernet networking interface, integrated-services digital network (ISDN) card, cable modem, satellite modem, or a modem to provide a data communication connection to a corresponding type of communications line, for example an Ethernet cable or a metal cable of any kind or a fiber-optic line or a telephone line. Network 270 broadly represents a local area network (LAN), wide-area network (WAN), campus network, internetwork, or any combination thereof. Communication interface 260 may comprise a LAN card to provide a data communication connection to a compatible LAN, or a cellular radiotelephone interface that is wired to send or receive cellular data according to cellular radiotelephone wireless networking standards, or a satellite radio interface that is wired to send or receive digital data according to satellite wireless networking standards. In any such implementation, communication interface 260 sends and receives electrical, electromagnetic, or optical signals over signal paths that carry digital data streams representing various types of information.

[0201] Network link 265 typically provides electrical, electromagnetic, or optical data communication directly or through at least one network to other data devices, using, for example, satellite, cellular, WiFi, or BLUETOOTH technology. For example, network link 265 may provide a connection through a network 270 to a host computer 250.

[0202] Furthermore, network link 265 may provide a connection through network 270 or to other computing devices via internetworking devices and / or computers that are operated by an Internet Service Provider (ISP) 275. ISP 275 provides data communication services through a world-wide packet data communication network represented as Internet 280. A server computer 255 may be coupled to Internet 280. Server 255 broadly represents any computer, data center, virtual machine, or virtual computing instance with or without a hypervisor, or computer executing a containerized program system such as DOCKER or KUBERNETES. Server 255 may represent an electronic digital service that is implemented using more than one computer or instance and that is accessed and used by transmitting web services requests, uniform resource locator (URL) strings with parameters in HTTP payloads, API calls, app services calls, or other service calls. Computer system 205 and server 255 may form elements of a distributed computing system that includes other computers, a processing cluster, server farm or other organization of computers that cooperate to perform tasks or execute applications or services. Server 255 may comprise one or more sets of instructions that are organized as modules, methods, objects, functions, routines, or calls. The instructions may be organized as one or more computer programs, operating system services, or application programs including mobile apps. The instructions may comprise an operating system and / or system software; one or more libraries to support multimedia, programming or other functions; data protocol instructions or stacks to implement TCP / IP, HTTP or other communication protocols; file format processing instructions to parse or render files coded using HTML, XML, JPEG, MPEG or PNG; user interface instructions to render or interpret commands for a graphical user interface (GUI), command-line interface or text user interface; application software such as an office suite, Internet access applications, design and manufacturing applications, graphics applications, audio applications, software engineering applications, educational applications, games or miscellaneous applications. Server 255 may comprise a web application server that hosts a presentation layer, application layer and data storage layer such as a relational database system using structured query language (SQL) or no SQL, an object store, a graph database, a flat file system or other data storage.

[0203] Computer system 205 can send messages and receive data and instructions, including program code, through the networks), network link 265 and communication interface 260. In the Internet example, a server 255 might transmit a requested code for an application program through Internet 280, ISP 275, local network 270 and communication interface 260. The received code may be executed by processor 210 as it is received, and / or stored in storage 215, or other non-volatile storage for later execution.

[0204] The execution of instructions as described in this section may implement a process in the form of an instance of a computer program that is being executed and consisting of program code and its current activity. Depending on the operating system (OS), a process may be made up of multiple threads of execution that execute instructions concurrently. In this context, a computer program is a passive collection of instructions, while a process may be the actual execution of those instructions. Several processes may be associated with the same program; for example, opening up several instances of the same program often means more than one process is being executed. Multitasking may be implemented to allow multiple processes to share processor 210. While each processor 210 or core of the processor executes a single task at a time, computer system 205 may be programmed to implement multitasking to allow each processor to switch between tasks that are being executed without having to wait for each task to finish. In an embodiment, switches may be performed when tasks perform input / output operations, when a task indicates that it can be switched, or on hardware interrupts. Timesharing may be implemented to allow fast response for interactive user applications by rapidly performing context switches to provide the appearance of concurrent execution of multiple processes simultaneously. In an embodiment, for security and reliability, an operating system may prevent direct communication between independent processes, providing strictly mediated and controlled interprocess communication functionality.

Claims

CLAIMS1 . Computer-implemented method (100) for providing at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, characterized in that it comprises: a step of defining (135) at least one value representative of a post-fermentation food or beverage product performance target, a step of generating (140) at least one blend digital identifier by a trained generative machinelearning model using as input each value representative of a post-fermentation food or beverage product performance target, a step of providing (145) at least one of the generated at least one blend digital identifier.

2. Computer-implemented method (100) according to claim 1 , wherein, during the step of generating(140), a value representative of a quantity is generated for at least one strain represented by a strain digital identifier, said strain digital identifier being associated with at least one generated blend digital identifier.

3. Computer-implemented method (100) according to claims 1 or 2, which comprises the steps of: providing (125) an original set of exemplar blend digital identifiers, said set of exemplar blend digital identifiers being representative of materialised blends comprising at least two existing strains for food or beverage product fermentation, said blend digital identifiers being associated with at least one value representative of an empirically measured post-fermentation food or beverage product performance, to form a training set and training (130) a generative machine-learning model using the training set, wherein the generative machine-learning model is trained to associate at least one value representative of an empirically measured post-fermentation food or beverage product performance with at least one strain digital identifier, wherein said trained generative machine-learning model is used during the step of generating.

4. Computer-implemented method (100) according to claim 3, which comprises: a step of culturing (105) a food or beverage product with a blend of at least two strains, a step of empirically measuring (110) at least one value representative of a post-fermentation food or beverage product performance,a step of storing (115), in a database, said at least one value representative of an empirically measured post-fermentation food or beverage product performance in association with at least two strain digital identifier representatives of the at least two strains used during the step of culturing, wherein said database is used during the step of providing a training set.

5. Computer-implemented method (100) according to claims 3 or 4, wherein at least one value representative of a post-fermentation food or beverage product performance target is at least one of: a time to reach (“TTR”) a potential of hydrogen of a predetermined limit value, a potential of hydrogen after an elapsed predetermined fermentation time, a value representative of a texture of the post-fermentation food or beverage product, a value representative of a storage of the post-fermentation food or beverage product, a value representative of a cell count at the end of fermentation and / or during shelf-life, and / or a value representative of a concentration of at least one chemical compound involved in the fermentation process.

6. Computer-implemented method (100) according to claims 1 to 5, wherein at least one value representative of post-fermentation food or beverage product performance target is at least two different points representative of the evolution of the potential of hydrogen over time.

7. Computer-implemented method (100) according to claims 1 to 6, wherein the trained generative machine-learning model is a conditional variational auto-encoder (“CVAE”).

8. Computer-implemented method (100) according to claims 1 to 7, wherein during the step of generating (140): a quantity of at least one additional physical additive to be added to the blend is generated, said additive being represented by an additive digital identifier, and / or at least one culturing operational parameter of the fermentation process is generated.

9. Computer-implemented method (100) according to claims 1 to 8, which further comprises a step of sending (150) a digital command representative of an instruction of materialising at least one blend corresponding to at least one blend digital identifier provided.

10. Computer-implemented method (100) according to claim 9, which further comprises a step of materialising (155) at least one blend corresponding to at least one blend digital identifier provided.

11. Computer-implemented method (120) to train a generative machine learning model to provide at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, characterized in that it comprises the steps of: providing (125) an original set of exemplar blend digital identifiers, said set of exemplar blend digital identifiers being representative of materialised blends comprising at least two existing strains for food or beverage productfermentation, each existing strain for food or beverage product fermentation being represented by a strain digital identifier, said blend digital identifiers being associated with at least one value representative of an empirically measured postfermentation food or beverage product performance, to form a training set, and training (130) a generative machine-learning model using the training set, wherein the generative machine-learning model is trained to associate at least one value representative of an empirically measured post-fermentation food or beverage product performance with at least one strain digital identifier.

12. Computer program product (200) characterized in that it comprises instructions which upon execution by a computer cause the computerto execute the method (100) according to any one of claims 1 to 11.

13. Computer-readable storage medium (215) storing programming instructions which upon execution by a computer cause the computer to execute the method (100) according to any one of claims 1 to 11.

14. Device (200) for providing at least one blend digital identifier, representing a materialisable blend, associated with at least two strain digital identifiers, each strain digital identifier being representative of an existing strain for food or beverage product fermentation, characterized in that it comprises: a means of defining at least one value representative of a post-fermentation food or beverage product performance target, a means of generating at least one blend digital identifier by a trained generative machinelearning model using as input each value representative of a post-fermentation food or beverage product performance target, a means of providing at least one of the generated at least one blend digital identifier.

Images