Method for generating a set of manufacturing data for a cosmetic composition to be modified, and associated devices

Abstract

Method for generating a set of manufacturing data for a cosmetic composition to be modified, and associated devices The present invention relates to a method for generating a method for generating at least one set of manufacturing data of a cosmetic composition, said method comprising a step of: - obtaining a set of manufacturing data and one or more manufacturing data items to be modified in the set obtained, - computing a similarity coefficient between each manufacturing data item to be modified and a plurality of candidate manufacturing data, - selecting, for each manufacturing data item to be modified, at least one manufacturing data item according to the similarity coefficients computed, to obtain at least one selected data item, - generating a set of manufacturing data, to obtain a generated set wherein each manufacturing data item to be modified is replaced by a selected manufacturing data item.

Description

[0001]Method for generating a set of manufacturing data for a cosmetic composition to be modified, and associated devices CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to FR Patent Application No.2407749, filed 15 July 2024, which is hereby incorporated by reference. TECHNICAL FIELD OF THE INVENTION The present invention relates to a method for generating a set of manufacturing data allowing the formulation, preparation and / or manufacture of a cosmetic composition. It also relates to a method for manufacturing a cosmetic composition. It also relates to the devices involved in the preceding methods. TECHNOLOGICAL BACKGROUND OF THE INVENTION The formulation of environmentally-friendly cosmetic compositions, i.e. the design and development of which account for environmental issues, is becoming a major concern to help meet global challenges. Hence, it has become essential to propose more sustainable compositions and / or preparation methods and / or raw materials thereby allowing addressing these environmental issues. In particular, in this context, it is becoming important to replace certain materials by alternatives having a better environmental footprint, particularly by reducing the use of raw materials obtained from petrochemicals. One objective is to provide eco-responsible materials, in particular derived from green chemistry, in order to limit the environmental impact of the composition (and complementarily of its primary and / or secondary packaging) from the manufacture thereof to the end-of-life thereof, in particular thanks to materials with a good biodegradability profile and / or derived from renewable sources. The formulation of novel cosmetic compositions may also be the result of new regulations. For example, following European directives of 2015, producing lipstick or nail polish containing cobalt nitrate has been prohibited. Reformulation may also be required because a raw material may be acceptable in one legislation but not in another. More broadly, it may also be desirable to modify a cosmetic composition in the aim of optimizing its manufacturing cost, optimizing its industrialization, or according to the availability of its constituent ingredients. It may also be desirable to modify it to improve and develop its properties for consumers. It may also be useful when it comes to being able to ensure the availability of ingredients or raw materials of certain cosmetic compositions. SUMMARY OF THE INVENTION There is therefore a need for a method which is easy to implement and makes it possible to generate a set of manufacturing data of a cosmetic composition addressing a specific context, in particular a reduced environmental footprint. To this end, an object of the description is a method for generating a method for generating at least one set of manufacturing data, the set of manufacturing data comprising manufacturing data of a cosmetic composition, said method being implemented by a computing device and comprising a generation phase, the generation phase comprising a step of: - obtaining a set of manufacturing data and one or more manufacturing data items to be modified in the set obtained, - computing a similarity coefficient between each manufacturing data item to be modified and a plurality of candidate manufacturing data, - selecting, for each manufacturing data item to be modified, at least one manufacturing data item from the plurality of candidate manufacturing data according to the computed similarity coefficients, to obtain at least one selected data item, - generating at least one set of manufacturing data, to obtain at least one generated set, each generated set being the set obtained wherein each manufacturing data item to be modified is replaced by a respective selected manufacturing data item. According to other advantageous aspects of the invention, the generation method comprises one or more of the following features, considered separately or in any technically- feasible combination: - each manufacturing data item is associated with several values, each value being a quantification of a respective property of the manufacturing data item, the similarity coefficient of a manufacturing data item being dependent on at least one value of the manufacturing data item. - a property is a measurable property. - a property is a qualitatively assessable property. - during the computing step, the computing device (12) computes at least one similarity coefficient by applying the following formula: Where: •^^(^,^) denotes the similarity coefficient between the manufacturing data item Aand the manufacturing data item B, •^^^^,^ denotes the i-th manufacturing data item to be modified, i being a non-zerointeger, •^^^^^, ^ denotes the j-th candidate manufacturing data item, j being a non-zerointeger, •^ denotes the sum of the weights applied to the p properties• ^^ denotes the weight applied to the p-th property ( ^^ = 1 by default)• ^ is the number of properties,|^^^^^ ‒ ^^^^^, ^ |• ^^(^^^ , ^ ) = ^ ^^^,^ ^^^^, ^ ^^o ^mod^^ : value of the p-th property of ^^^^,^ , where p is an integer between1 and n, o^^^^^, ^^ : value of the p-th property of ^^^^^, ^ ,o ^^ : this value is at least equal to the difference between the 5th and 95thpercentile of the value distribution of the p-th property. -for a specific manufacturing data item to be modified, the selected candidatemanufacturing data item is the deletion of the specific manufacturing data item, each generated set then being the obtained set wherein the specific manufacturing data item is deleted and each other manufacturing data item to be modified possibly is replaced by a respective selected manufacturing data item. -the cosmetic composition is part of a cosmetic product, the manufacturing databeing chosen from a list consisting of the data relating to the cosmetic composition, the data relating to a packaging of the cosmetic product and the data relating to the used cosmetic product manufacturing technique. -the manufacturing data are chosen from a list of raw materials, a list of molecularstructures and a list of characteristics. The description also relates to a method for generating a set of manufacturing data, the set of manufacturing data comprising manufacturing data of a cosmetic composition capable of meeting at least one acceptability criterion, said method being implemented by a computing device and comprising an inference phase, the inference phase comprising a step of: -obtaining a set to be completed, the set comprising manufacturing data of acosmetic composition, and optionally at least one constraint parameter to be fulfilled by the chemical composition, and -applying a technique to the set to be completed and optionally the at least oneconstraint parameter to obtain a completed set, the completed set comprising manufacturing data of a cosmetic composition capable of meeting said at least one acceptability criterion, the technique comprising the use of at least one machine learning model, the at least one machine learning model being capable of completing a set to be completed. According to other advantageous aspects of the invention, the generation method comprises one or more of the following features, considered separately or in any technically- feasible combination: -the learning model has been trained to complete a set to be completed withlearning on at least one database compiling cosmetic compositions meeting said at least one acceptability criterion and optionally cosmetic compositions not meeting said at least one acceptability criterion. -the method furthermore comprises a training phase of the model on a trainingdatabase during which a merit function is used, the merit function favoring models capable of generating completed sets which are not part of the training database and meeting the at least one acceptability criterion. -the manufacturing data are a list of raw materials or ingredients, optionallyaccompanied by their respective proportions in the cosmetic composition. -during the obtaining step, at least one constraint parameter is obtained, aconstraint parameter being a completeness parameter defining a number of missing manufacturing data items in the set to be completed. -during the application phase, the model is used only once, the completenessparameter corresponding to a number of missing data items being equal to 1. -a constraint parameter is an overall completeness parameter corresponding to anumber of missing data items greater than 1, the application phase comprising several iterations, each iteration comprising the application of the model to the set generated in the preceding iteration with a completeness parameter for said iteration corresponding to a number of missing data items equal to 1, the model being applied to the first iteration on the set to be completed.- the model comprises an encoder and / or a decoder.- The model is a masked predictive transformer model.- the cosmetic composition is part of a cosmetic product, the manufacturing databeing chosen from a list consisting of the data relating to the cosmetic composition, the data relating to a packaging of the cosmetic product and the data relating to the used cosmetic product manufacturing technique. -the method further comprises a phase of updating the model by retraining themodel using the sets used during the interference phase. The description also relates to a method for manufacturing a cosmetic composition, the method being implemented by a manufacturing system, the manufacturing method comprising: - a phase of implementing a method for generating at least one set of manufacturing data, the generation method being as described above, and - a phase of manufacturing a cosmetic composition from each set generated during the implementation phase. The description also describes a computing device configured to generate at least one set of manufacturing data, the set of manufacturing data comprising manufacturing data of a cosmetic composition, and to: - obtain a set of manufacturing data and one or more manufacturing data items to be modified in the set obtained, - compute a similarity coefficient between each manufacturing data item to be modified and a plurality of candidate manufacturing data, - select, for each manufacturing data item to be modified, at least one manufacturing data item from the plurality of candidate manufacturing data according to the similarity coefficients computed, to obtain at least one selected data item, - generate at least one set of manufacturing data, to obtain at least one generated set, each generated set being the set obtained wherein each manufacturing data item to be modified. The description also relates to a system for manufacturing a cosmetic composition, the manufacturing system comprising: - a computing device as described before, and - a manufacturing device capable of manufacturing a cosmetic composition from the set of manufacturing data obtained by the computing device. The description also describes a method for predicting a set of characteristics associated with a list of raw materials, said method being implemented by a prediction device and comprising an inference phase, the inference phase comprising a step of: - obtaining a list of properties associated with the list of raw materials, and - applying a technique to the list of properties to obtain the characteristics of the raw material and / or the list of raw materials, the technique comprising the use of at least one Bayesian network, said Bayesian network being determined to link characteristics of a raw material with properties. According to other advantageous aspects of the invention, the prediction method comprises one or more of the following features, considered separately or in any technically- feasible combination: -the method further comprises the application of a function f linking the value of acharacteristic with a distribution of a node and with the list of properties according to the following equation: ^^ = ^(^(^|^^,1, ^, ^^,^) )Where: o^^ : value of the characteristic i,o ^(^|^^,1, ^, ^^,^) : probability of the node i knowing the n parent nodes of i.- arcs and distributions of the Bayesian network are determined by an expert group or using known data. - the list of characteristics comprises the set of characteristics for which the associated probability distribution is greater than an associated prediction threshold. - the method further comprises an update phase during which the user can update the Bayesian network according to expert opinions and known data. - the method comprises the application of an unsupervised machine learning algorithm to expert opinions and known data to determine the Bayesian network. -a characteristic is a measurable property.- a characteristic is a qualitatively assessable property.The description also describes a method for selecting one or more lists of candidate raw materials to replace a first list of raw materials and comprising the following steps: -determining a list of properties associated with the first list of raw materials andan associated sensory attribute, -computing characteristics of the cosmetic composition by implementing aprediction method as described above,- applying a machine learning model to determine a plurality of lists of candidateraw materials according to the sensory attribute for which the characteristics are similar to the computed characteristics, and -selecting one or more lists of raw materials from the plurality of lists of candidateraw materials according to a selection criterion associated with a threshold assigned to each of the characteristics. The description also describes a device for predicting a set of characteristics associated with a list of raw materials, said prediction device being configured to -obtain a list of properties associated with the list of raw materials, and- apply a technique to the list of properties to obtain the characteristics of the rawmaterial and / or the list of raw materials, the technique comprising the use of at least one Bayesian network, said Bayesian network being determined to link characteristics of a raw material with properties. BRIEF DESCRIPTION OF THE FIGURES The invention will become more apparent upon reading the following description, given solely as a non-limiting example, and made with reference to the drawings wherein: oFigure 1 is a schematic representation of a manufacturing system comprising acomputing device and a manufacturing device, oFigure 2 is a flowchart of an example of implementation of a method forgenerating a set of manufacturing data, oFigure 3 is a block representation of an example of a model used by thegeneration method of Figure 2, oFigure 4 is a block representation of another example of a model used by thegeneration method of Figure 2, oFigure 5 is a block representation of yet another example of a model used by thegeneration method of Figure 2, oFigure 6 is a flowchart of an example of implementation of another method forgenerating a set of manufacturing data, oFigure 7 is a schematic representation of an example of latent space allowing abetter understanding of the method of Figure 6, oFigure 8 is a flowchart of an example of implementation of a method formanufacturing a cosmetic composition, and oFigure 9 is a flowchart of an example of implementation of a prediction method. DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS Description of the manufacturing system In Figure 1, a system 10 for manufacturing a cosmetic product is shown schematically. The manufacturing system 10 is capable of manufacturing the cosmetic product. The concept "cosmetic product" is defined in the section relating to the manufacturing data of a cosmetic product. The manufacturing system 10 comprises a computing device 12 and a manufacturing device 14. The computing device 12 is capable of implementing a method for generating a set of manufacturing data making it possible to manufacture a cosmetic product, or at least a part thereof, that is to say at least one cosmetic composition. Several examples of generation methods will be described hereinafter. These methods are computer-implemented methods. Further information on the manufacturing data which are part of the set is contained in the corresponding section of the description. The computing device 12 is a computer comprising one or more electronic components such as: one or more processors with one or more cores represented collectively by a processor, one or more graphics processing units (GPU), a hard drive, a random access memory (RAM), a display interface or an input / output interface. It will be appreciated that the computer may for example be implemented in the form of a desktop computer, an on-board computer in a vehicle, a tablet or a smartphone. In the example of Figure 1, the computing device 12 comprises a processing unit 16 comprising and a memory 18 associated with the processing unit 16. The processing unit 16 is an electronic circuit designed to handle and / or transform data represented as electronic or physical quantities in registries of the computer and / or memories into other similar data corresponding to physical data in memories of registries or other types of display devices, transmission devices or storage devices. As specific examples, the processing unit 16 is embodied in the form of a programmable logic component, such as an FPGA (Field Programmable Gate Array), or an integrated circuit, such as an ASIC (Application-Specific Integrated Circuit). Alternatively, when the method is embodied in the form of one or more software programs, i.e. in the form of a computer program, also referred to as computer program product, it is furthermore capable of being saved on a computer-readable medium, not shown. For example, the computer-readable medium is a medium able to store electronic instructions and to be coupled to a bus of a computer system. For example, the readable medium is an optical disk, a magneto-optical disk, a ROM memory, a RAM memory, any type of non-volatile memory (for example FLASH or NVRAM) or a magnetic card. A computer program comprising software instructions is then stored on the readable medium. The manufacturing device 14 is capable of manufacturing the cosmetic product from the set of manufacturing data generated by the processing unit 16. Description of a generation method from a set of data to be completed The operation of the computing device 12 is now described with reference to Figure 2 which is a flowchart illustrating an example of implementation of a generation method. The generation method is a method aimed at generating a set of manufacturing data of a cosmetic composition capable of meeting at least one acceptability criterion from a set of manufacturing data to be completed. Incompleteness of the set of manufacturing data to be completed is conveyed by a completeness parameter forming a context input parameter which, as will be seen below, may be implicit or explicit. The completeness parameter conveying the incompleteness of the set of manufacturing data can particularly be expressed in the form of a number of missing raw materials and / or a total number of raw materials to be reached in the completed composition. It is understood that it can alternatively or additionally be expressed in the form of a constraint parameter, such as a price, performance, environmental compatibility constraint, inducing the search for additional raw materials to complete the set. This constraint parameter can be different from the acceptability criterion or correspond to the acceptability criterion. In this description, the term "manufacturing data" refers to data allowing formulating and / or preparing the cosmetic product and could also be expressed as "preparation data". Hereinafter in the description, the set that is generated will be referred to as "the completed set" (meeting the completeness parameter) and the input set will be referred to as "set to be completed" (not meeting the completeness parameter). More specifically, in the example that will be described, it is assumed that the set to be completed is a list of constituent raw materials of a cosmetic composition which, in principle, does not meet the acceptability criterion. The set to be completed is a list from which one or more raw materials is / are implicitly or explicitly missing. In particular, the set to be completed is a list for which the number of constituent raw materials is less than the number of raw materials sought after applying the completeness parameter. For example, when the completeness parameter is the total number of expected raw materials, then the total number of raw materials of the composition to be completed is less than the completeness parameter. In the case where the completeness parameter is the number of raw materials missing or to be added, then said completeness parameter added to the number of raw materials of the composition to be completed corresponds to the completed number of raw materials of the sought composition and to be generated by the model described below. The completeness parameter can also be a range of values giving the model the freedom to add a number of raw materials between a lower limit and an upper limit. It can also be considered to provide only a lower limit value or an upper limit value. Where appropriate, it is also possible to provide for a default value. In addition, it is assumed that the completed set is a completed list of raw materials, capable of meeting at least one acceptability criterion, it being understood that this example could be transposed to any type of manufacturing data as described in the corresponding section, namely the section entitled "Description of the different sets of manufacturing data that can be considered". For example, the set to be completed may be an empty set (with no raw material). The aim of the method is therefore that of generating the data making it possible to manufacture a cosmetic composition comprising, in a physiologically acceptable medium, a plurality of raw materials. A raw material is a set of one or more ingredients used in the manufacture of cosmetic products. These raw materials can be of natural or synthetic origin and play various roles in cosmetic formulations, ranging from active agents to texture agents. For example, a raw material may be glycerin, elastin or hyaluronic acid. Hereinafter in this description, a raw material is defined by the set of its constituent ingredients, a set of molecular structures of its constituent ingredients, a set of concentrations of its constituent ingredients, a set of properties of its constituent ingredients or a set of characteristics. By "physiologically acceptable", it should be understood a medium compatible with keratin materials. By "keratin materials", it should be understood the skin, the mucosa and / or the skin appendages. Preferably, the keratin materials are the skin, in particular the skin of the face, the mucosa such as the lips, and / or the skin appendages such as the eyelashes. By way of non-limiting example, such a cosmetic composition is a hair treatment or care product (e.g. a shampoo), a skin treatment or care product (e.g. a moisturizing cream, sun cream, anti-aging cream), a make-up product (e.g. a lipstick, mascara, foundation, or gloss). In the example described, the generation method comprises a learning phase P20, an inference phase P30 and an update phase P40. This is merely a simple example, it being understood that in general the inference P30 and update P40 phases are implemented successively to always improve the model used. It can also be noted that the implementation of the update phase P40 is not mandatory. The learning phase P20 may be carried out offline, i.e. by a computer other than the computing device 12, and, preferably, before the use of the computing device 12. During the learning phase P20, a machine learning model M is trained via a machine learning algorithm A to complete a set to be completed according to a completeness parameter defining, directly or indirectly, a number of missing manufacturing data items in the set to be completed. A machine learning algorithm A is an algorithm which teaches / trains a model M to find patterns and / or general input-output relationships automatically from training (or learning) data. Upon completion of the learning, the model M would be able to make similar inferences and / or predictions. In mathematical terms, the model M could be considered as an equation mapping inputs to output. Machine learning algorithms can be categorized according to the types of tasks to be performed (classification, regression, generation), the learning approach used (supervised, unsupervised, semi-supervised, self-supervised and reinforced) and the algorithmic theory (neural network, tree model, probabilistic model, polynomial model, Bayesian model, SVM, etc.). The learning algorithm applied in this method relates to self-supervised and reinforced learning for a generation task. It is based on the neural network theory, or more specifically the Transformer theory for training and inference / generation. The model M at the output of the learning phase is a Transformer type neural network. A neural network comprises an ordered sequence of neural layers, each of which takes its inputs from the outputs of the preceding layer. More specifically, each layer comprises neurons taking their inputs from the outputs of the neurons of the preceding layer, or from the input variables for the first layer. Alternatively, more complex neural network structures may be considered with a layer that can be connected to a layer farther away than the immediately previous layer. An operation, i.e. a type of processing, to be performed by said neuron within the corresponding processing layer is also associated with each neuron. Each layer is connected to the other layers by a plurality of synapses. A synaptic weight is associated with each synapse, and each synapse forms a link between two neurons. It is often a real number, which takes both positive and negative values. In some cases, the synaptic weight is a complex number. Each neuron is capable of performing a weighted sum of the value(s) received from the neurons of the preceding layer, each value then being multiplied by the respective synaptic weight of each synapse, or connection, between said neuron and the neurons of the preceding layer, then applying an activation function, typically a non-linear function, to said weighted sum, and outputting from said neuron, in particular to the neurons of the following layer which are connected thereto, the value resulting from the application of the activation function. The activation function allows introducing a non-linearity into the processing performed by each neuron. The sigmoid function, the hyperbolic tangent function, the Heaviside function are examples of activation functions. As on optional complement, each neuron is also capable of also applying a multiplying factor, also referred to as bias, to the output of the activation function, and the value output from said neuron is then the product of the bias value and the value derived from the activation function. A convolutional neural network is also sometimes referred to by the acronym CNN. In a convolutional neural network, each neuron of the same layer has exactly the same connection pattern as its neighboring neurons, but at different input positions. The connection pattern is referred to as a convolution kernel. A fully connected neuron layer is a layer wherein the neurons of said layer are each connected to all the neurons of the preceding layer. Such a type of layer is more commonly called "fully connected", and sometimes referred to as "dense layer". Alternatively, the model M applies a linear regression, a logistic regression, a decision tree, a random forest, a principal component analysis (PCA), a Naive Bayes algorithm or a K-nearest neighbors (KNN) algorithm Alternatively, the model M is a support vector machine or a K-Means Clustering model. By "machine learning", it should herein be understood that the model M is taught to carry out a task using a learning algorithm A run on a computer and applied to training data. The learning of the M model on a task is performed via learning algorithms A which are an optimization process relative to a well-defined metric. By way of example, the optimization can be minimization of the least squares error between predicted and actually measured values or maximization of the reward in the case of reinforced learning, on training data. The task to be carried out here is the generation of the completed set of raw materials from a set to be completed or a subset of the completed set. The model M can thus be seen as a generative or predictive model. It is also assumed that a training database has been established for this purpose. This learning database contains examples of data that the model M can read repetitively to understand the patterns or implicit relationships present in the data. The generation of such a database varies according to the type of Transformer and the learning approach adopted. For a Transformer that only contains the encoder and is learnt via a learning approach called "MLM" (Masking Language Modeling), each learning example consists of a list to be completed some raw materials of which are masked and another completed list all the raw materials of which are set out. Each pair of data items of the base can be generated from an actual cosmetic composition of which the list of raw materials is known and masking one or more raw materials from the list. An actual cosmetic composition is implicitly considered chemically valid and minimally completed. During training of an encoder type Transformer, the list to be completed will be submitted to the input of the model M. The input data information will be propagated in the different layers of the model for processing up to the output layer. The output layer will predict the masked raw materials at the input. The predicted raw materials will be compared with the corresponding raw materials in the completed list by computing an error. The error will be used to readjust the parameters in the model M via the back propagation mechanism in order to reduce this error. This mechanism is performed repetitively in order to progressively reduce the overall error on masked raw material prediction. For a Transformer containing only a decoder, each learning example consists of a completed list of raw materials in a cosmetic composition with a 'START' token added at the beginning of the list and the same completed list with an 'END' token added at the end. Each pair of data items of the base can be generated starting from an actual cosmetic composition the list of raw materials of which is known and by adding the 'START' token at the beginning of the list or the 'END' token at the end of the list, respectively. An actual cosmetic composition is considered chemically valid and minimally completed. During training of such a model containing only a decoder, the list containing the 'START' token will be injected into a self-attention layer by the 'VALUE', 'KEY' and 'QUERY' inputs. The self-attention layers encode the information in the list and propagate it towards the output layer to predict the list containing the 'END' token. The weights in the self-attention layers which are recomputed for each token in the list containing the 'START' token will be masked by a so-called 'CAUSAL' mask. This 'CAUSAL' masking could mimic an autoregressive generation mechanism. The prediction error is computed by comparing the sequence of raw materials generated in such an autoregressive way and the subset list containing the 'END' token. The error will be used to readjust the parameters in the model M via the back propagation mechanism in order to reduce this error. This mechanism is performed repetitively in order to progressively reduce the overall error on masked raw material prediction. For a Transformer containing an encoder and a decoder, each learning example consists of a list to be completed with a subset of raw materials removed from a completed cosmetic composition list, and two lists containing an additional subset of raw materials in the same completed cosmetic composition list. One of the complementary lists is added an additional 'START' token at the beginning of the list and the other is added an additional 'END' token at the end of the list. Each learning data triplet can be generated by splitting a completed actual cosmetic composition list and adding the 'START' token at the beginning of the additional list or 'END' at the end of the additional list, respectively. An actual cosmetic composition is implicitly considered chemically valid and minimally completed. During training of a Transformer containing an encoder and a decoder, the list to be completed of a subset of raw materials is injected into an encoder which encodes it into a vector which represents all the essential information for subsequent prediction. This encoded vector will be injected into the decoder by the 'KEY' and 'VALUE' inputs of the second self-attention layer. The additional list containing the 'START' token will be injected into the same self- attention layer by the 'QUERY' input, after having been encoded by another preceding self- attention layer. The information encoded by the second self-attention layer will be propagated in the remaining layers up to the output layer to predict the complementary list containing the 'END' token. The weights in the self-attention layers, which are recomputed for each token in the list containing the 'START' token, will be masked by a so-called 'CAUSAL' mask. This 'CAUSAL' masking could mimic an autoregressive generation mechanism. The prediction error is computed by comparing the sequence of raw materials generated in such an autoregressive way and the subset list containing the 'END' token. The error will be used to readjust the parameters in the model M via the back propagation mechanism in order to reduce this error. This mechanism is performed repetitively in order to progressively reduce the overall error on masked raw material prediction. Via this simple example, it is apparent that it is possible to generate a large database because a single composition can make it possible to generate a plurality of pairs for training. Any training technique can then be used to obtain the model. For example, the learning is supervised or unsupervised learning or self-supervised learning that has been applied in the process. Other techniques may be considered such as reinforcement learning techniques used in the process to encourage the model to propose sets of manufacturing data allowing manufacturing a cosmetic product which is not part of the learning database and meeting at least one acceptability criterion. According to one embodiment, a merit function (also referred to as loss function) is used, the merit function favoring models capable of generating sets of manufacturing data considered as "completed" allowing manufacturing a cosmetic product which is not part of the database and meeting at least one acceptability criterion. More specifically, learning is carried out by estimating weights or scores associated with a reward function. In this example, the generation of known data is neutral in relation to the reward function. On the other hand, generating data which are part of a list of data to be excluded creates a negative weight or score, whereas generating unknown data creates a positive weight or score. By producing an average, the model M determines an associated acceptability criterion making it possible to validate a future generation or not. For other types of models M like those based on the K-nearest neighbors (KNN) algorithm, the training comprises storing the training data, computing Euclidean distances between the test points and all training points, selecting the near neighbors, and predicting by classification or regression (see Figure 6 in particular). During the inference phase P30, the computing device 12 applies the learned model to input data comprising a composition to be completed and a completeness parameter. The completeness parameter can be explicit, for example by directly providing the expected total number of raw materials or the number of missing raw materials to be added. The completeness parameter can also be defined by default or implicit and determined from the data of the composition to be completed using a tag or masking token in the data of said composition to be completed. According to one particular embodiment, the completeness parameter is inferred from an additional constraint parameter, such as a price constraint, for example. According to the example described, the inference phase P30 comprising an obtaining step E301 and an application step E302. During the obtaining step E301, the computing device 12 obtains a set of manufacturing data to be completed and the completeness parameter. For example, the computing device 12 reads this list and the completeness parameter in the memory 18. Furthermore, in some embodiments, the user can enter a completed list and select at least one raw material to be excluded with which a MASK token will subsequently be associated. In such a case, the number of MASK tokens determines the completeness parameter (e.g. exact, minimum or maximum number of raw materials to be added). During the application step E302, the computing device 12 applies a technique to the set to be completed to obtain a completed set. Possibly associated with a probability of said completed set meeting the acceptability criterion of the model. The completed sets for which the probability of meeting the criterion is greater than a determined threshold (>0.8 for example) will be presented to the user. The technique here consists in applying the model M to the set to be completed to obtain a completed set according to the completeness parameter and the acceptability criterion. Several topologies for the model M may be considered. According to a first example, the model M is applied only once. In this example, the model M comprises at least a part of a transformer, more specifically the encoder. As a particular example, the model M uses a BERT ("Bidirectional Encoder Representations from Transformers") transformer model. The MPT model is a neural network comprising several layers shown in Figure 3. As a general rule, the model M has several layers, of which two sets of layers play an advantageous role. The first set is a multi-head attention set. This first set allows the model to estimate the connection between the input elements and to extract relationships between near elements and far elements with the same efficiency. The second set is a feed forward point-to-point neural network. Such a neural network is more often referred to as "Position-Wise Feed-Forward Network". The second set serves here to help change the representation and capture more information about the context. A particular example of such an example of model is now described with reference to Figure 3. In the example shown, the model M comprises 4 encoders. Each encoder comprises an "input embedding" layer, a "multi-head attention" layer which comprises two "addition / normalization" sub-layers, a "feed forward" layer and a "classification" layer. The "input embedding" layer is a layer in charge of converting presented tokens into integer identifiers in a continuous vector format. The "multi-head attention" layer continues to encode different tokens previously generated with weightings (also referred to as attention weights). These weights are computed for each token and represent implicit relational information between the tokens with respect to the final prediction. The "addition / normalization" layers consist in concatenating each original token (before passage via the "multi-head attention" layer) at the output of each multi-head attention sub-layer, to thus enrich the number of informative features used for the final prediction. A normalization is then performed to stabilize the learning process. The "feed forward" layer is composed of several neural sub-layers or each of the sub-layers are interconnected. The neurons of each of said sub-layers receive weighted inputs from the neurons of the preceding layer and transmit their outputs to the next layers. At each activated layer, a scalar product is applied between the weights associated with each neuron and their inputs before application of an activation function. All the training of such a neural network consists in converging the values of said weights to match the theoretical outputs of the training data with the actual outputs. The outputs of the "feed forward" layer are "contextualized embeddings". The "classification layer" allows transforming the normalized contextualized embeddings concatenated with the tokens entering the feed forward layer into task-specific predictions. Typically, such predictions are text classification, named entity recognition, or responses to questions. In our example, they consist of raw materials for manufacturing a cosmetic composition. According to a second example, the model M is applied several times with a re- injection at the input of all or part of the completed sets obtained at the output of the model M. The re-injection is possibly accompanied by an update of the completeness parameter (addition of a masked raw material in the re-injected data for an implicit determination of the updated completeness parameter or explicit update of the completeness parameter). This means that at a first computing iteration, the model M is applied to the set to be completed and outputs a first intermediate set then, at a second iteration, the model is applied to the first intermediate set to output a second intermediate set and so on up to a final iteration during which the application of the model M makes it possible to obtain the completed set. According to one embodiment, the number of iterations of the application corresponds to the number of missing manufacturing data items in the set to be completed. In such a case, at each iteration, the model M adds a manufacturing data item to the set provided at the input of the model M at each iteration. According to one embodiment, the overall completeness parameter is greater than 1 missing raw material, for example 3 missing raw materials, each iteration using an intermediate completeness parameter corresponding to 1 missing raw material. Thus, after 3 iterations, the completed composition is obtained. In such a case, the model M is a masked predictive transformer model. Such a model is often referred to by the abbreviation MPT, which refers to "Masked Predictive Transformer". The model M is a neural network comprising several layers shown in Figure 4. In the shown example, the model M comprises an "input embedding" layer, two "multi-head attention" layers, three "addition / normalization" layers, a "feed forward" layer, a "linear" layer, a "softmax" layer and an "argmax" layer. The linear layer is a layer used in neural networks which applies a linear transformation (or matrix multiplication with a transformation matrix) to input data using weights and biases. The application of this linear layer makes it possible to transform the data received at its input into either a lower dimensionality for dimensional reduction, or higher dimensionality for more complex tasks. The softmax layer is a layer used in neural networks which applies a softmax function (or a normalized exponential function) to input data in real numbers, in order to convert them into a format which could represent the probability distribution on a number of choices. It is useful for assigning a probability to each possible prediction response. The argmax layer always follows the softmax layer. The application of this layer is simply that of outputting the element with the highest value in the probability vector generated by the softmax layer. This element will be considered as the final response for the prediction. According to a third example, the model M is applied several times with a re-injection at the input of the second output sub-model of the model M. This means that at a first computing iteration, the model M is applied to the set to be completed and outputs at least one first intermediate set then, at a second iteration, the model is applied to one of the first intermediate sets to output at least one second intermediate set and so on up to a final iteration during which the application of the model M makes it possible to obtain the completed set. According to one embodiment, the number of iterations of the application corresponds to the number of missing manufacturing data items in the set to be completed. In such a case, at each iteration, the model M adds a manufacturing data item to the set provided at the input of the model M at each iteration. In such a case, the model M is a masked predictive transformer model. Such a model is often referred to by the abbreviation MPT, which refers to "Masked Predictive Transformer". The model M is a neural network comprising several layers shown in Figure 5. The model M comprises a first sub-model and a second sub-model. The first sub-model comprises an "input embedding" layer, a "multi-head attention" layer, two "addition / normalization" layers and a "feed forward" layer. The output of the first sub-model is then applied to the input of a second "multi-head attention" layer of the second sub-model comprising an "input embedding" layer, a first "multi-head attention" layer, three "addition / normalization" layers, the second "multi-head attention" layer, a "feed forward" layer, a "linear" layer, a "softmax" layer and an "argmax" layer. During the update phase P40, the completed set obtained is, where applicable after verifying whether it meets the acceptability criterion or not, integrated with training data and the model M is retrained using the training data obtained. For example, the weights associated with each sub-layer of the "feed forward" layer are determined by the computing device 12 in order to match the model outputs to the training data. The generation method thus makes it possible to obtain a completed set. Through these examples, it is apparent that the machine learning model can also take into account one or more additional input parameters making it possible to determine the acceptability criterion / criteria. These input parameters are constraint parameters to be fulfilled by the cosmetic composition (and therefore indirectly by the completed set). As an example of additional input data, mention may be made of the number of missing raw materials, in which case the constraint parameter is a completeness parameter or a quantification of a performance that the cosmetic product should achieve. In the case where the constraint parameter is a completeness parameter, it could be considered to implement the preceding embodiments as follows. During the application phase, the model is used only once, the completeness parameter corresponding to a number of missing data items equal to 1. In the iterative embodiment, a constraint parameter is an overall completeness parameter corresponding to a number of missing data items greater than 1 and the application phase comprises several iterations, each iteration comprising the application of the model to the set generated at the preceding iteration with a completeness parameter for said iteration corresponding to a number of missing data items equal to 1, the model being applied to the first iteration on the set to be completed. In the case where the constraint parameter is a quantification of a performance to be achieved by the cosmetic product, this quantification makes it possible to define an acceptability criterion. More generally, the acceptability criterion is inferred from the available information, either only from the constraint parameter(s), particularly when the set to be completed is empty (with no raw material), or from the set to be completed, in particular in the absence of constraint parameters or from both, i.e. the raw materials of the set to be completed and the parameters. In an extreme case, the acceptability criterion can thus be reduced to a biologically acceptable formula for the part of the body intended to receive it. Where appropriate, the term "raw material" could include a packaging or member for applying the cosmetic composition, the packaging and the member then being intended to form, together, a cosmetic product. Thus, based on the same principle described above, it is possible to provide, to a model, the data of a cosmetic composition considered as incomplete, no longer in relation to a missing chemical raw material, but in relation to a missing application member or packaging to form a final cosmetic product, this final cosmetic product forming the completed cosmetic composition in the sense of the embodiments described above. Of course, it is possible to combine these acceptability criteria to generate the most satisfactory completed set possible, the model being trained accordingly in accordance with the acceptability criteria to be targeted. Description of a generation method starting from a set of data to be modified The operation of the computing device 12 is now illustrated with reference to Figure 6 which is a flowchart illustrating another example of implementation of another generation method. The generation method according to Figure 6 comprises a determination phase P50 and a generation phase P60. During the determination phase P50, a list of properties and / or molecular structures associated with each of the raw materials is determined. During the generation phase P60, the computing device 10 obtains one or more sets to be completed. The generation phase P60 comprises an obtaining step E601, a computing step E602, a selection step E603 and a generation step E604. During the obtaining step E601, the computing device 12 obtains a list of raw materials and one or more raw materials to be modified in the set obtained. For example, the user has just entered the list of raw materials to be modified via the input interface which the computing device 12 reads in the memory 18. During the computing step E602, the computing device 12 computes a similarity coefficient between each raw material to be modified and a plurality of candidate raw materials. According to one particular example, each raw material is associated with several property values. The term property is to be understood here in the broadest sense as comprising physically measurable properties but also qualitative properties assessable on an appropriate rating scale. More specifically, the term property refers to physical, chemical, functional or biological properties. For example, a physical property defines the structure of the raw material such as its state, solubility or viscosity. For example, a chemical property chemically defines the raw material via its pH, associated carbon chain length or oxidation. For example, a functional property defines different practical abilities of the raw material such as its hydrating, anti-aging or emollient ability. For example, a biological property defines a biological ability of the raw material as its antibacterial, healing or soothing ability. Each value is a quantification of a respective property of the raw material, so that providing all of the values of the properties of a raw material could be viewed as providing a signature of the raw material. The computed similarity coefficient is then dependent on at least one value of a property of the raw material. Several examples of such computing can be given. In one embodiment with properties for which the value is quantifiable, the computer 12 applies the following equation: Where: •^^(^,^) denotes the similarity coefficient between the material A and the materialB, •^^^^,^ denotes the i-th raw material to be modified, i being a non-zero integer,• ^^^^^, ^ denotes the j-th candidate raw material, j being a non-zero integer,• ^ denotes the sum of the weights applied to the p properties• ^^ denotes the weight applied to the p-th property ( ^^ = 1 by default)• ^ is the number of properties,|^^^^^ ‒ ^^^^^, ^ |• ^^(^^^ , ^ ) = ^ ^^^,^ ^^^^, ^ ^^o ^mod^^ : value of the p-th property of ^^^^,^ , p being an integer between 1and n, o^^^^^, ^^ : value of the p-th property of ^^^^^, ^ ,o ^^ : this value is at least equal to the difference between the 5th and 95thpercentile of the value distribution of the p-th property. •If the value of the property is qualitatively measurable: According to one example, the number of properties n is equal to the number of different properties of the raw materials of the set given during the obtaining step E601. According to one variant, the computing step E602 is implemented differently. In this variant, the computing step E602 comprises a search step and a computing step. During the search step, the computing device 12 searches a plurality of reference lists. Typically, a reference list is composed of sets having undergone, or not, modifications in their associated raw material lists. The reference lists are stored in a database via a shared domain data set (SDDS, standing for Shared Domain DataSet) allowing storing a large number of data items while defining associated characteristics such as the source of said data. Typically, such a structure makes it possible to store more than one million reference lists in the database. The reference lists relate to cosmetic compositions which may have been modified in the past. Thus, each list is associated with a plurality of lists of raw materials. In the example described, the reference list comprises N iterations and therefore N lists of associated distinct raw materials. Furthermore, the reference list comprises the raw material to be modified but all the lists of distinct raw materials generated after the J-th iteration no longer comprise it (J being less than N). For example, the search step is performed using an SQL query applied to the database. Thus, the computing device 12 identifies the N-J lists of associated distinct raw materials not comprising the raw material to be modified. During the computing step E602, the computing device 12 computes a similarity coefficient between the N-J lists of associated distinct raw materials and the J lists of raw materials containing the raw material to be modified. For example, the computing device 12 applies a function f such that:^^^,^ = ^(^,^)Where: o^^(^,^) denotes the similarity coefficient between the set to be completed and theB-th list of associated distinct materials (^ ≤ ^ ≤ !) ,o ^ : number of iterations,o X: number of raw materials not shared between the two lists.Alternatively, the computing step E602 is implemented differently. In this variant, the computing step E602 comprises a conversion sub-step, a first computing sub-step, a projection sub-step and a second computing sub-step. In the conversion sub-step, the molecular structure of the raw material is converted into the form of molecular identifiers. For example, such identifiers take the form of SMILES, InchiKey or Molecular fingerprint structures. During the first computing sub-step, the computing device 12 computes a plurality of molecular properties from the molecular identifiers. For example, the molecular properties include the molar mass, a partition coefficient or a number of chemical bonds. During the projection sub-step, the computing device 12 identifies the molecular properties linked with each other in order to represent them in a reduced-dimension projection space (see Figure 7). For example, the computing device 12 minimizes the dimensions of the projection space by identifying the dimensions linked with molecular properties as linear compositions of other properties. Typically, the computing device 12 can apply a principal component analysis PCA of such a step. Principal component analysis is a statistical analysis method which allows summarizing the information contained in a set of functional data, thereby allowing producing so-called principal component functions of minimal dimension which reproduce maximum information on the studied data. In other words, functional principal component analysis is a projection of input information in an optimal projection base. Finally, during the second computing step, the computing device 12 computes the similarity coefficient by applying a clustering algorithm in the projection space of dimension M and computing the similarity coefficient according to a distance criterion. For example, the similarity coefficient is zero for two raw materials which are not in the same cluster. Advantageously, such a characteristic makes it possible to replace the at least one raw material to be excluded by a new raw material providing one or more priority properties for the operator. Alternatively, a method for predicting characteristics of the raw material according to said properties can be used. This device is described in the final section of the present application. During the selection step E603, the computing device 12 selects, for each raw material to be modified, at least one raw material from the plurality of candidate raw materials. Each raw material thus selected is hereinafter referred to as selected raw material. The selection is performed according to the similarity coefficients computed during the computing step E602. According to one simple example, the selected raw materials are the raw materials of the plurality of candidate raw materials having the highest similarity coefficients. In other examples, the raw materials are selected according to similarity coefficients and a set of criteria associated with each of the candidate raw materials. Typically, the set of criteria comprises an economic or environmentally friendly indicator such as biodegradability. During the generation step E604, the computing device 12 generates all the sets of raw materials wherein each raw material to be modified is replaced by a respective selected raw material. The generation method thus makes it possible to obtain one or more completed sets of raw materials. This particularly makes it possible to exclude a raw material by replacing it with one or more others or delete it to obtain a new cosmetic composition having the sought properties. Such a method then makes it possible to generate manufacturing data useful for the preparation and manufacture of innovative cosmetic compositions adapted to specific constraints. Advantageously, the implementation of such a method is easy, in particular relatively rapid. Furthermore, this method also allows access to formulation, improvement or replacement levers for any person manufacturing cosmetic products, from beginners to experienced users or even experts. Applications of the preceding generation methods The set obtained by one of the preceding generation methods is useful for numerous applications. According to the example described, it can be used in a method for preparing and manufacturing a cosmetic product by the manufacturing system 14. The manufacturing method comprises an implementation phase P70 and a manufacturing phase P80. During the implementation phase P70, the computing device 12 implements one of the preceding generation methods. For example, the computing device 12 implements the inference phase P40 of the generation method according to the embodiment of Figure 2 or the generation phase P60 of the generation method according to the embodiment of Figure 6. The manufacturing phase P80 aims to manufacture the cosmetic product corresponding to the set obtained after implementation during the implementation phase P70. For example, the manufacturing device 14 takes the generated set as an input and controls its constituent elements so that these elements carry out the manufacture of the cosmetic product corresponding to the set generated by the generation method. Other manufacturing methods may be considered. For example, during a configuration step, the operator can select a packaging and an embodiment to be fulfilled in the form of data and include them in the manufacturing data. The manufacturing technique refers to a set of criteria to be met during the manufacture of the cosmetic composition like an order of incorporation, temperature, mixing elements (speed, time, impellers and / or blades), breaking down into phases or pre-mixing. In particular, the manufacturing technique may be expressed in the form of method unitary operations. Description of the different sets of manufacturing data that can be considered The generation methods of Figures 2 and 6 have so far been described for a particular example of a list of raw materials for illustrative purposes. However, in a general case, the manufacturing data can be of any type. Depending on the cases, the manufacturing data are chosen from a list consisting of data relating to the composition, data relating to the packaging, particularly to an application member of the cosmetic composition, and data relating to the manufacturing technique used. The manufacturing data make it possible, depending on the cases, to obtain a cosmetic composition or a cosmetic product. A cosmetic composition is composed of a list of raw materials assigning a list of properties to said composition. By "cosmetic product", it should be understood all of a cosmetic composition, a packaging, comprising particularly an application member. The manufacturing method denotes a set of unitary operations to be fulfilled during the manufacture of the cosmetic composition such as an order of incorporation, temperature, humidity or applied force. The packaging denotes the means for storing the cosmetic composition. Alternatively, the packaging comprises an application member or element of the associated cosmetic composition such as a mascara lash brush or a foundation brush A data item relating to the composition is a data item representative of the chemical composition of the formula. The list of raw materials or ingredients is a particular example of data relating to the composition. Another example of data relating to the composition can be the raw material concentration in said composition. The representation of the raw material can differ according to the embodiments. By way of example, the raw material can be represented by a chemical formula, an identifier, a trade name or a type (emulsifier or fatty substance) and its proportion in the composition can be expressed as a mass or volume percentage, detailed linearly or with preparations Instead of a raw material, it is possible to consider ingredients, which are part of the raw materials, where the ingredients can similarly be represented in different ways (formula, identifier or others). The raw material or ingredient content is another example of data relating to the composition. A data item relating to the used manufacturing technique is a data item allowing characterizing the steps of a manufacturing method. An example of data relating to the manufacturing technique is a sequence of unitary method operations. An example of data relating to the manufacturing technique is the order in which the ingredients or the raw material are inserted in the manufacture, tank placement temperature, introduction and mixing times, speeds and mechanical forces applied. Description of a method for predicting characteristics As explained above, the preceding generation methods use knowledge on lists of raw materials. It is therefore useful to be able to estimate or predict the characteristics of a raw material and / or a list of raw materials For this purpose, a prediction method can be implemented via a prediction device. The same observations as for the computing device are applicable here for the prediction device and are therefore not repeated. An example of implementation of the method for predicting characteristics of a list of raw materials is now described The prediction method is a method aimed at estimating a set of characteristics associated with a raw material. Alternatively, the method estimates a set of characteristics associated with a plurality of raw materials forming a cosmetic composition. Thus, hereinafter in the description, a raw material also denotes a set of raw materials associated with a list of previously determined properties from the lists of properties of the set of raw materials. The term characteristic here is to be understood in the broadest sense as comprising measurable characteristics but also qualitative characteristics assessable in an appropriate way, for example on a rating scale according to defined assessment criteria. The characteristics are information associated with the raw material such as ensured functions, technical characteristics, physicochemical properties or contributions to sensory benefits. The characteristics of a raw material are defined by all the technical effects induced by its properties. For example, a characteristic of a raw material is its tackiness, its acid resistance or its ability to stabilize the oily phase. Such a list of characteristics also comprises a value associated with each of the characteristics which can be quantitative or qualitative. The prediction method according to Figure 9 comprises a determination phase P90 and a prediction phase P100. During the determination phase P90, a list of properties associated with the raw material is determined. The prediction phase P100 comprises an obtaining step E1001, a computing step E1002 and a prediction step E1003. During the obtaining step E1001, the computing device 12 obtains a list of properties. For example, the user manually enters the list of properties of the raw material via the input interface which the computing device 12 reads in the memory 18. The list of properties of the raw material can also be retrieved automatically by the system, in particular from a suitable storage database. In another example, the computing device 12 receives the list of properties of the raw material by an electronic device via radio communication or wired connection. Furthermore, the user enters values associated with each of said properties. As for the list of properties, the associated values can also be retrieved automatically from a database. During the computing step E1002, the computing device 12 computes the list of characteristics using the list of properties and a causal Bayesian network. A Bayesian network is a model making it possible to determine an output value or a distribution of output values according to a plurality of input data influencing (or not) said output value. A Bayesian network is a graphical probabilistic model which represents a set of random variables and their conditional dependencies via a directed acyclic graph. It is used to model uncertain knowledge in various fields such as the links between characteristics and properties of a cosmetic composition. To create such a Bayesian network, it is necessary to determine, in advance of use, a set of variables representing the nodes of the graph, a set of arcs linking the nodes representing conditional dependencies and probabilities for each of the nodes. Each node corresponds to a characteristic, the value of which is then dependent on the network inputs (values of the properties entered by the user). Each node is associated with a parent influencing its probability distribution. The inputs (properties) are also represented by nodes without parents. The arcs and probabilities associated with each of the nodes allow this dependency. Indeed, the arcs and probabilities are determined by expert opinion from a group of experts and represent the dependencies between characteristics and properties, these are known as elicited probabilities. Alternatively, the arcs and probabilities are determined from known data. Typically, known data are data linking a property with a historically known or recently discovered characteristic. The values assigned to each of the probabilities are modifiable over time, for example if new knowledge makes it possible to refine the probabilities determined by the expert group. Thus, an arc linking a property with a characteristic means that the property influences said characteristic. An intermediate node is present on each of the arcs making it possible to represent the weight (influence) of the property on said characteristic. In addition, intermediate nodes can make it possible to represent a cumulative effect of several properties on a characteristic linking several weights together. For example, the list of characteristics comprises the set of characteristics for which the associated probability distribution is greater than an associated prediction threshold. During the prediction step E1003, the prediction device predicts the list of characteristics and their associated values. For example, the prediction device applies a function f linking the distribution with a value of the characteristics and with the properties such that: ^^ = ^(^(^|^^,1, ^, ^^,^) )Where: o^^ : value of the characteristic i,o ^(^|^^,1, ^, ^^,^) : probability distribution of the node i knowing the n parentnodes of i, In one embodiment, during an optional update phase, it is possible to update the Bayesian network via the prediction device. The term "update" means adding or modifying associated nodes, probabilities and arcs. For example, the user enters new nodes, new arcs, and new associated probabilities via the prediction device and based on expert opinions of expert groups. Alternatively, the Bayesian network, new nodes, new arcs, and new probabilities are determined by applying one or more unsupervised learning algorithms to the known data. The unsupervised learning algorithm(s) are part of the group: clustering algorithms, dimensionality reduction algorithm, density modeling algorithm, or neural network algorithm. De tels algorithmes d^apprentissage non supervisés sont présentés dans l^article intitulé « Learning Bayesian Networks with the bnlearn R Package » publié le 10 juillet 2010. Advantageously, this "hybrid" approach between the expert opinion and the result of one or more learning algorithm(s) allows increasing the accuracy and the perenniality over time of the results of the causal Bayesian model. Advantageously, it is possible to obtain the list of properties of a raw material from these characteristics by pulling up the Bayesian network graph. The term "pull up" means "determine" and is performed by applying an optimization algorithm or Bayes' theorem. Typically, the optimization algorithm is a genetic algorithm. The prediction method can be used in a method for selecting one or more lists of raw materials to replace a first list of raw materials implemented by the computing device. Such a selection method comprises a determination step, a computing step, and a selection step. During the determination step, a list of properties associated with the first list of raw materials and an associated sensory attribute are determined. A sensory attribute refers to a known and intrinsic characteristic of a cosmetic product, perceptible by a human user via at least one of their five senses: sight, smell, taste, touch or hearing. These quantifiable or qualifiable attributes contribute to the overall sensory experience perceived during use or interaction with said product. For example, the user manually enters the list of properties and the sensory attribute via the input interface which the computing device 12 reads in the memory 18. During the computing step, the computing device 12 computes the characteristics of the cosmetic composition using a prediction method as described above. During the selection step, the user selects a list of raw materials from the plurality of lists of candidate raw materials via the computing device 12. Alternatively, the computing device 12 selects one or more lists meeting a set of selection criteria dependent on the sensory attribute. For example, a selection criterion is a threshold assigned to each of the characteristics. Advantageously, such a method makes it possible to select one or more lists of raw materials having similar characteristics to the input list.

Claims

CLAIMS 1. A method for generating at least one set of manufacturing data, the set ofmanufacturing data comprising manufacturing data of a cosmetic composition, said method being implemented by a computing device (12) and comprising a generation phase, the generation phase comprising a step of: - obtaining a set of manufacturing data and one or more manufacturing data items to be modified in the set obtained, - computing a similarity coefficient between each manufacturing data item to be modified and a plurality of candidate manufacturing data, - selecting, for each manufacturing data item to be modified, at least one manufacturing data item from the plurality of candidate manufacturing data according to the computed similarity coefficients, to obtain at least one selected data item, - generating at least one set of manufacturing data, to obtain at least one generated set, each generated set being the set obtained wherein each manufacturing data item to be modified is replaced by a respective selected manufacturing data item.

2. The generation method according to claim 1, wherein each manufacturingdata item is associated with several values, each value being a quantification of a respective property of the manufacturing data item, the similarity coefficient of a manufacturing data item being dependent on at least one value of the manufacturing data item.

3. The generation method according to claim 2, wherein a property is ameasurable property.

4. The generation method according to claim 2 or 3, wherein a property is aqualitatively assessable property.

5. The generation method according to any one of claims 2 to 4, wherein, duringthe computing step, the computing device (12) computes at least one similarity coefficient by applying the following formula:Where: •^^(^,^) denotes the similarity coefficient between the manufacturing data item Aand the manufacturing data item B, •^^^^,^ denotes the i-th manufacturing data item to be modified, i being a non-zerointeger, •^^^^^, ^ denotes the j-th candidate manufacturing data item, j being a non-zerointeger, •^ denotes the sum of the weights applied to the p properties• ^^ denotes the weight applied to the p-th property ( ^^ = 1 by default)• ^ is the number of properties,|^^^^^^ ‒ ^^^^^, ^ |) = ^^ ^^value of the p-th property of ^ , where p is an integern,: value of the p-th property of ^^^^,o ^^ : this value is at least equal to the difference between the 5th and 95thpercentile of the value distribution of the p-th property.

6. The generation method according to any one of claims 1 to 5, wherein, for aspecific manufacturing data item to be modified, the selected candidate manufacturing dataitem is the deletion of the specific manufacturing data item, each generated set then beingthe obtained set wherein the specific manufacturing data item is deleted and each other manufacturing data item to be modified possibly is replaced by a respective selected manufacturing data item.

7. The generation method according to any one of claims 1 to 6, wherein thecosmetic composition is part of a cosmetic product, the manufacturing data being chosen from a list consisting of the data relating to the cosmetic composition, the data relating to a packaging of the cosmetic product and the data relating to the used cosmetic product manufacturing technique.

8. The generation method according to any one of claims 1 to 7, wherein themanufacturing data are chosen from a list of raw materials, a list of molecular structures and a list of characteristics.

9. A method for manufacturing a cosmetic composition, the method beingimplemented by a manufacturing system (10), the manufacturing method comprising:- a phase of implementing a method for generating at least one set of manufacturing data, the generation method being according to any one of claims 1 to 8, and - a phase of manufacturing a cosmetic composition from each set generated during the implementation phase.

10. A computing device (12) configured to generate at least one set ofmanufacturing data, the set of manufacturing data comprising manufacturing data of a cosmetic composition, and to: - obtain a set of manufacturing data and one or more manufacturing data items to be modified in the set obtained, - compute a similarity coefficient between each manufacturing data item to be modified and a plurality of candidate manufacturing data, - select, for each manufacturing data item to be modified, at least one manufacturing data item from the plurality of candidate manufacturing data according to the similarity coefficients computed, to obtain at least one selected data item, - generate at least one set of manufacturing data, to obtain at least one generated set, each generated set being the set obtained wherein each manufacturing data item to be modified.

11. A manufacturing system (10) for a cosmetic composition, the manufacturingsystem (10) comprising: - a computing device (12) according to claim 10, and - a manufacturing device (14) capable of manufacturing a cosmetic composition from the set of manufacturing data obtained by the computing device (12).

Images