Residual Semi-Recurrent Neural Networks

Abstract

To provide an artificial neural network that improves prediction accuracy.SOLUTION: Residual semi-recurrent neural networks (RSNN) is configured to receive both time invariant input and time variant input data to generate one or more time-series predictions. The time invariant input can be processed by a multilayer perceptron of the RSNN. The output of the multilayer perceptron can be used as an initial state for a recurrent neural network unit of the RSNN. The recurrent neural network unit can also receive time invariant input, and process the time invariant input with the time variant input to generate an output. The outputs of the multilayer perceptron and the recurrent neural network unit can be combined to generate the one or more time-series predictions.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No. 62 / 824,895, filed March 27, 2019, and European Patent Application Publication No. 19305611.6, filed May 13, 2019, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure relates to artificial neural networks. [Background technology]

[0003] An artificial neural network (ANN) is a computing system loosely inspired by biological neural networks. An ANN is a framework for one or more machine learning algorithms to work together to process complex data inputs. ANNs can be trained to perform tasks such as image recognition, where an ANN can be trained to identify several features (e.g., animals, vehicles, road signs, etc.) in an image. ANNs typically contain a collection of connected nodes called artificial neurons. Each connection can transmit a signal from one artificial neuron to another. An artificial neuron can process a signal and then output the results to additional connected artificial neurons.

[0004] In typical embodiments, the signals at the connections, sometimes called edges, between artificial neurons are real, and the output of each artificial neuron is calculated according to a nonlinear function of the sum of its inputs. Artificial neurons and edges typically have weights that can be adjusted while the ANN is being trained. ANNs have been used for a variety of tasks, including computer vision, speech recognition, machine translation, and medical diagnosis. More recently, ANNs have been used as predictive models to predict outcomes based on several inputs. For example, ANNs are used in pharmacokinetic modeling to predict how drugs interact with the human body (e.g., the time course of drug absorption, distribution, metabolism, and excretion). Summary of the Invention [Means for solving the problem]

[0005] In at least one aspect of the present disclosure, a neural network is provided. The neural network includes a multilayer perceptron (MLP) configured to receive external data of a first type and process the external data of the first type to generate an MLP output. The neural network includes a recurrent neural network (RNN) unit. The RNN unit is configured to receive at least a portion of the MLP output and external data of a second type. The RNN unit is configured to process at least a portion of the MLP output and at least the second type of external data to generate an RNN output. The MLP and RNN units are together configured to generate one or more time series predictions based at least in part on the RNN output and the MLP output.

[0006] The MLP may include an input layer, at least one hidden layer, and an output layer. The RNN units may include at least one hidden RNN layer. The RNN units may include long short-term memory RNN units. The RNN units may include regular RNN units. The RNN units may include gated recurrent units.

[0007] The first type of external data may include time-invariant data, and the second type of external data may include time-varying data.

[0008] The neural network is configured to combine the MPL output and the RNN output to generate a residual output. One or more time series predictions are generated based at least in part on the residual output. The one or more time series predictions may include concentration values ​​of the drug in plasma as a function of time after administration of a dose of the drug.

[0009] In at least one other aspect of the present disclosure, a method is provided. The method includes receiving, by a multilayer perceptron (MLP), external data of a first type. The method includes processing, by the MLP, the external data of the first type to generate MLP outputs. The method includes receiving, by a recurrent neural network (RNN) unit, at least a portion of the MLP outputs and external data of a second type. The method includes processing, by the RNN unit, at least a portion of the MLP outputs and the external data of the second type to generate RNN outputs. The method includes generating one or more time series predictions based at least in part on the RNN outputs and the MLP outputs.

[0010] The first type of external data may include time-invariant data, and the second type of external data may include time-varying data.

[0011] The method may further include combining the MPL output and the RNN output to generate a residual output. One or more time series predictions are generated based at least in part on the residual output. The one or more time series predictions may include concentration values ​​of the drug in plasma as a function of time after administration of a dose of the drug.

[0012] The RNN units may include long short-term memory RNN units. The RNN units may include regular RNN units. The RNN units may include gated recurrent units.

[0013] Embodiments of the present disclosure may include one or more of the following advantages: The described neural networks can process time-varying and time-invariant data in a computationally efficient manner to generate accurate time series forecasts compared to traditional neural networks, such as recurrent and convolutional neural networks; the output of a particular neural network layer is combined with the output of a successive neural network layer to increase forecast accuracy compared to traditional neural networks; and, compared to traditional neural networks, the described neural networks can better capture the relationship between time-varying and time-invariant data and how this relationship affects the resulting time series forecasts.

[0014] These and other aspects, configurations, and embodiments may be expressed as methods, apparatus, systems, components, program products, means or steps for performing a function, or otherwise.

[0015] These and other aspects, configurations, and embodiments will become apparent from the following description, including the claims. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 depicts a residual semi-recurrent neural network, in accordance with one or more embodiments of the present disclosure. [Figure 2A] FIG. 1 depicts a recurrent neural network unit, in accordance with one or more embodiments of the present disclosure. [Figure 2B] FIG. 1 depicts an expanded recurrent neural network unit in accordance with one or more embodiments of the present disclosure. [Figure 3] FIG. 1 depicts the architecture of a residual semi-recurrent neural network, in accordance with one or more embodiments of the present disclosure. [Figure 4]1 is a flowchart depicting a method for generating one or more time series predictions using time-varying and time-invariant data, in accordance with one or more embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0017] The use of artificial neural networks (ANNs) as predictive models to forecast future outcomes based on several external data sources is becoming more common. For example, pharmacokinetic (PK) modeling, a key step in drug development, models how a drug interacts with the human body (e.g., the time course of a drug's absorption, distribution, metabolism, and excretion). Some PK modeling techniques use time-invariant data, such as a patient's baseline characteristics (e.g., age and gender), and time-varying data, such as the patient's drug dose level, to generate a PK curve, which shows the drug's concentration in plasma over a time interval. Traditionally, differential equations borrowed from the field of fluid mechanics have been used to model drug interactions with the human body. While differential equations can be useful for describing the underlying physical processes that occur in the body after receiving a drug dose, they may not fully account for patient-to-patient heterogeneity. Therefore, predictions obtained using these differential equations for individual patients can be inaccurate, whereas predictions at the population level can be accurate.

[0018] A common approach to improving the predictive accuracy of pharmacokinetic modeling is typically to use feedforward neural networks (e.g., convolutional neural networks). However, despite improving predictive accuracy compared to traditional differential equation approaches, feedforward neural networks may not be able to reveal correlations between pharmacokinetic measurements at multiple time points in the same individual. This is because traditional feedforward networks typically only process information unidirectionally (e.g., from the input layer through the hidden layer to the output layer). Therefore, feedforward networks only consider the current input and have no concept of temporal order.

[0019] Recurrent neural networks (RNNs) have been shown to handle time-varying data better because they contain nodes with connections that form a directed graph along a time series, allowing the RNN to process input sequences using internal memory. However, when using traditional RNNs to process time-invariant data, the RNN may process the time-invariant data in a manner similar to time-varying data. This can create a large amount of unnecessary duplicated data, which can degrade predictive performance, especially when using small sample sizes. Therefore, artificial neural networks adapted to efficiently process both time-invariant and time-varying data are desirable for generating time-based predictions with improved accuracy compared to traditional approaches.

[0020] The present disclosure provides a residual semi-recurrent neural network configured to process both time-invariant and time-varying data in an efficient manner that improves prediction accuracy compared to conventional RNN and feedforward networks. The residual semi-recurrent neural network includes a multilayer perceptron (MLP) for receiving and processing the time-invariant data. The residual semi-recurrent neural network also includes an RNN unit for processing the time-varying data. Outputs from the MLP and RNN units are used by the neural network to generate one or more time series predictions. For example, when used for PK modeling, data related to patient baseline characteristics are provided to the MLP, and data related to dose levels of a drug administered to the patient at sequential time points are provided to the MLP. is provided to the RNN. The output of the MLP is used as the initial state of the RNN. The outputs of the RNN and MLP are also combined to generate a Pk curve, which shows the relationship between the drug concentration in a patient's plasma and time after a single dose of the drug is administered to the patient.

[0021] In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent that the present disclosure may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to avoid unnecessarily obscuring the present disclosure.

[0022] In the figures, a particular arrangement or order of schematic elements, such as those representing devices, modules, instruction blocks, and data elements, is shown for ease of explanation. However, those skilled in the art should understand that the particular order or arrangement of schematic elements in the figures does not imply that a particular order or sequence of operations, or separation of operations, is required. Furthermore, the inclusion of a schematic element in a drawing does not imply that such element is required in all embodiments, or that the configuration represented by such element is not included in or combined with other elements in some embodiments.

[0023] Furthermore, when a connecting element, such as a solid or dashed line or arrow, is used in the drawings to illustrate a connection, relationship, or association between two or more other schematic elements, the absence of such a connecting element does not imply that the connection, relationship, or association cannot exist. In other words, some connections, relationships, or associations between elements may not be shown in the drawings so as not to obscure the disclosure. Additionally, for ease of illustration, a single connecting element may be used to represent multiple connections, relationships, or associations between elements. For example, when a connecting element represents communication of signals, data, or instructions, those skilled in the art will understand that such element represents one or more signal paths (e.g., buses) to affect the required communication.

[0024] Reference will now be made in detail to the embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the various embodiments described. However, it will be apparent to those skilled in the art that the various embodiments described may be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments.

[0025] Below are described several features, each of which can be used independently of one another or with any combination of the other features. However, any individual feature may not address all of the problems discussed above, or may only address one of the problems discussed above. Some of the problems discussed above may not be completely solved by any of the features described herein. Even if a heading is provided, data related to a particular heading may not be found in the section with that heading, but may be found elsewhere in this specification.

[0026] As used herein, the term process (processing, processes, etc.) in connection with a given neural network layer / unit refers to the known neural network process of applying learned weight functions / activation functions corresponding to each of the neurons included in the given neural network layer / unit to input data to generate an output for each neuron. As is known in the art, learned weight functions / activation functions are tuned / tuned by using a training data set to train the neural network and associating some features of the data set with some outcomes. It will be censored.

[0027] System Overview: 1 shows a diagram depicting a residual semi-recurrent neural network 100 in accordance with one or more embodiments of the present disclosure. The residual semi-recurrent neural network (RSNN 100) includes an MLP 110 and an RNN unit 120. The MLP 110 includes an input layer 111, a hidden layer 112, and an output layer 113. The RNN unit 120 includes a hidden RNN layer 122 and an output RNN layer 123. The MLP 110 is communicatively coupled to the RNN unit 120.

[0028] The input layer 111 of the MLP 110 includes a first input neuron 111a and a second input neuron 111b. The hidden layer 112 of the MLP 110 includes a first hidden neuron 112a, a second hidden neuron 112b, and a third hidden neuron 112c. The output layer 113 of the MLP 110 includes a first output neuron 113a and a second output neuron 113b. The layers 111, 112, and 113 of the MLP 110 are fully connected, meaning that each neuron in a given layer is connected to every neuron in the subsequent layer. For example, the first output neuron 113a is connected to the first hidden neuron 112a, the second hidden neuron 112b, and the third hidden neuron 112c.

[0029] Although MLP 110 includes one hidden layer 112, in some embodiments, MLP 110 includes more hidden layers (e.g., two hidden layers, three hidden layers, etc.) Each of layers 111, 112, 113 can include more or fewer neurons than the number shown in the illustrated embodiment.

[0030] The input layer 111 is configured to receive time-invariant data and send the time-invariant data to the hidden layer 112. Examples of time-invariant data may include values ​​associated with one or more patient baseline characteristics, such as the patient's age and / or gender. Each of the hidden nodes 112a, 112b, and 112c in the hidden layer 112 receives the time-invariant data and processes the time-invariant data to generate an output. The outputs of each of the hidden nodes 112a, 112b, and 112c in the hidden layer 112 are sent to a respective output node 113a and 113b. The output nodes 113a and 113b process the received outputs with their activation functions to generate MLP outputs.

[0031] The MLP 110 sends the MLP output to the RNN unit 120. The MLP output is received by the hidden RNN layer 122. The MLP output is then used as the initial state of the RNN unit 120. The hidden RNN layer 122 is also configured to receive time-varying input data. As described in more detail below with reference to FIG. 3, the RNN unit 120 is configured to sequentially process the received time-varying data to generate one or more time series predictions, using the MLP output from the MLP 110 as the initial state. For example, the MLP output may be obtained by processing input data related to a patient's baseline characteristics, and the time-varying data may be a series of cumulative doses of a drug infused into the patient. In this example, the output (time series prediction) of the RNN unit may be a Pk curve. In some embodiments, the RSNN 100 is configured to add the output of the MLP 110 to the output of the RNN unit 120 to generate the time series predictions.

[0032] Although RNN unit 120 is shown as having one hidden layer 122, in some embodiments, it has additional hidden RNN layers. In this example, MLP 110 is configured to provide its output to one or more of the additional hidden RNN layers. In some embodiments, RSNN 100 is configured to provide a single hidden RNN layer 122 to the MLP 110 and the RNN unit 122. 0. In these examples, MLP 110 is configured to provide its output to one or more of the additional fully connected layers. The amount of additional hidden RNN layers and fully connected layers is based on practical considerations such as processing needs, efficiency factors, and the type and amount of data to be processed.

[0033] 2A shows a diagram depicting an RNN unit 200, according to one or more embodiments of the present disclosure. The RNN unit 200 includes an input RNN layer 210, a hidden RNN layer 211, and an output RNN layer 212. The input RNN layer 210 includes a first RNN input neuron 210a, a second RNN input neuron 210b, and a third RNN input neuron 210c. In some embodiments, the input RNN layer 210 is the output layer of the immediately preceding MLP.

[0034] The hidden RNN layer 211 is configured to receive an output from the input RNN layer 210. As previously shown with reference to FIG. 1, the output from the input RNN layer is used as the initial state of the hidden RNN layer 211. The hidden RNN layer 211 includes a first recurrent neuron 211a and a second recurrent neuron 211b. Each of the RNN input neurons 210a, 210b, and 210c is coupled to a respective one of the recurrent neurons 211a and 211b. Each of the recurrent neurons 211a and 211b is configured to receive a first sequential data point, process the first sequential data point to generate a first output, and use the first output to process successive sequential data points. For example, assume the sequential data points are cumulative amounts of a drug infused into a patient. The first sequential data point may be the cumulative amount at a first time, and the second sequential data point may be the cumulative amount at a second time that occurs after the first time. Recurrent neurons 211a, 211b receive a first sequential data point, process the first sequential data point to generate a first output, and use information from the first output to process a second sequential data point.

[0035] The RNN output layer 212 is configured to receive the output from the hidden layer 211. The RNN output layer 212 includes a first RNN output neuron 212a, a second RNN output neuron 212b, a third RNN output neuron 212c, and a fourth RNN output neuron 212d. Each of the RNN output neurons 212a, 212b, 212c, and 212d is connected to each of the recurrent neurons 211a and 211b.

[0036] 2B shows a diagram depicting an unfolded recurrent neural network unit 300 in accordance with one or more embodiments of the present disclosure. As shown, the hidden RNN layer 321 is configured to receive and process a first sequential input 310a to generate a first output (S1). The hidden RNN layer 321 is also configured to receive a second sequential input 310b that follows the first sequential input 310a. The hidden RNN layer 321 processes the second sequential input 310b using the output (S1) associated with the first sequential input 310a to generate a second output (S2). Similarly, the hidden RNN layer 321 is configured to process a third sequential input 310c that follows the second sequential input 310b using the second output (S2) to generate a third output (S3).

[0037] 3 is a diagram depicting the architecture of an RSNN 400 in accordance with one or more embodiments of the present disclosure. The RSNN 400 includes an MLP 410 and a hidden RNN layer 422. The MLP 410 is configured to receive a time-invariant input 410. For example, the time-invariant input 410 may correspond to patient baseline characteristics such as age, gender, etc. The MLP 410 processes the invariant input 410 to generate an MLP output (S0), which is used as the initial state of the hidden RNN layer 422. The hidden RNN layer 422 receives a first sequential input 420a and processes the first sequential input 420a with the MLP output (S0) to generate a first RNN output (S1). The hidden RNN layer 422 then receives a second sequential input 420b. The hidden RNN layer 422 receives the second sequential input 420b and processes the second sequential input 420b using the first RNN output (S1) to generate a second RNN output (S2). The hidden RNN layer 422 then receives the third sequential input 420c and processes the third sequential input 420c using the second RNN output (S2) to generate a third RNN output (S3). The MLP output (S0) is then added 430 to the third RNN output (S3) to generate a time series prediction 440. In some embodiments, the sequential inputs correspond to the cumulative amount of drug infused at a given time. In some embodiments, the time series prediction 440 is a Pk curve.

[0038] Although the RNN units shown and described above are conventional RNN units, other types of RNN units may be used. In some embodiments, a long short-term memory (LSTM) RNN unit is used. An LSTM RNN unit is an RNN unit that includes a cell, an input gate, an output gate, and a forget gate. The cell stores a value over an arbitrary time interval, and the three gates control the flow of information. LSTM RNN units can minimize the effects of the vanishing gradient problem, which is known to be sometimes encountered when training conventional RNNs. In some embodiments, a gated recurrent unit (GRU) is used as the RNN unit. A GRU is an LSTM with a forget gate, but lacks an output gate, and therefore has fewer parameters than an LSTM. Compared to an LSTM, a GRU has been shown to exhibit better predictive performance when dealing with relatively small datasets.

[0039] 4 is a flowchart depicting a method 500 for generating one or more time series predictions using time-varying and time-invariant data according to one or more embodiments of the present disclosure. The method 500 includes receiving, by an MLP, external data of a first type (block 502); processing, by the MLP, the external data of the first type to generate an MLP output (block 504); receiving, by an RNN unit, at least a portion of the MLP output and external data of a second type (block 506); processing, by the RNN unit, at least a portion of the MLP output and the external data of the second type to generate an RNN output (block 508); and generating one or more time series predictions based at least in part on the RNN output and the MLP output.

[0040] At block 502, the MLP of the RSNN receives a first type of external data. In some embodiments, the first type of external data is time-invariant data. For example, the time-invariant data can be baseline patient characteristics.

[0041] At block 504, the MLP processes the first type of external data to generate the MLP output described above with respect to FIG.

[0042] At block 506, the RNN unit of the MLP receives at least a portion of the MLP output from the MLP. The RNN unit also receives a second type of external data. In some embodiments, the second type of external data is time-varying data. For example, the time-varying data can be a series of cumulative doses of a drug infused into a patient.

[0043] At block 508, the RNN unit processes the received portion of the MLP output together with the second type of extrinsic data to generate the RNN output described above with respect to FIG.

[0044] At block 510, the RSNN generates one or more time series predictions based at least in part on the RNN output and the MLP output. In some embodiments, the RSNN combines the RNN output and the MLP output to generate one or more time series predictions. The time series predictions may be, for example, Pk curves.

[0045] An example method for training an RSNN: Next, an exemplary method for training an RSNN for pharmacokinetic modeling is described.

number

number

number

[0046] The model is trained on a given instance i.

number

number

number

number

[0047] The baseline characteristics are provided to the multilayer perceptron (MLP) of the RSNN. In this case, the output of the MLP is used as the initial state of the conventional RNN unit of the RSNN. Time-varying data (e.g., the dose level assigned to a patient and at a given time) is provided to the conventional RNN unit. Because time-invariant inputs may have more influence on the Pk model than time-varying inputs, a shortcut connection is implemented between the MLP used to process the time-invariant inputs and the fully connected layer stacked on top of the recurrent layer. This shortcut provides a means to directly propagate information contained in the time-invariant inputs to the final output of the RNN unit.

[0048] Mathematically, an RSNN is defined as follows: Given an instance i, a time-varying input x i , and a series of time-varying inputs

number

number

number

number

[0049] To model pharmacokinetics, z it =d it Similarly, an MLP may contain multiple fully connected layers. Additionally, if an RNN unit has multiple recurrent layers, the output of the MLP may be the initial state of only the first recurrent layer, only the last recurrent layer, or all recurrent layers. Additionally, if multiple fully connected layers are stacked on top of a recurrent layer, the output of the MLP may be provided to only the first fully connected layer, only the last fully connected layer, or all fully connected layers. These different design decisions are considered hyperparameters.

[0050] Test Results: A dataset consisting of 65 patients was provided to an exemplary RSNN using gated recurrent neural networks (GRUs) as RNN units. For each patient, a set of observed PK values ​​(y i ), and two baseline characteristics: age and sex (x i ), as well as the cumulative amount of drug injected (d i ) was used. A GRU was used to concatenate the inputs and outputs of the RSNN at each time point in the dataset, and the results (e.g., Pk curves) of the training and validation sets were fitted.

[0051] The RSNN model accurately fit all PK curves in the validation set, including the two highly peaked PK curves seen in the training set, demonstrating that the RSNN can capture the complex nonlinear relationship between PK values ​​and the cumulative amount of infused drug, as well as baseline characteristics.

[0052] In the foregoing description, embodiments of the present disclosure have been described with reference to numerous specific details that may vary from embodiment to embodiment. Accordingly, the specification and drawings are not to be taken in a limiting sense. and should be considered exemplary. The sole and exclusive indication of the scope of the present disclosure and what applicants deem to be the scope of the present disclosure is the literal and equivalent scope of the set of claims originating from this application, in the specific form derived from such claims, including any subsequent amendments. Any definitions expressly set forth herein for terms contained in such claims shall govern the meaning of such terms as used in the claims. Additionally, when the term "further comprising" is used in the foregoing description or in the claims that follow, what follows this phrase may be an additional step or entity, or a sub-step / sub-entity of a previously recited step or entity.

Claims

1. 1. A method implemented by one or more computers, the method comprising: receiving data characterizing an object; and processing data characterizing the subject to generate one or more predictions characterizing the subject; The data characterizing the subject are: (i) time-varying data characterizing a subject, including respective time-varying data for each time point in a series of time points; (ii) includes time-invariant data characterizing the subject that is the same for each time point in the series of time points; Processing data characterizing the subject includes: processing the time-invariant data using a multi-layer perceptron (MLP) to generate an MLP output; Initializing hidden states of a recurrent neural network based on the MLP outputs; initializing hidden states of a recurrent neural network based on the MLP output, and then processing the time-varying data using the recurrent neural network to generate one or more predictions that characterize the subject; For each time point in the series of time points, processing the time-varying data and the recurrent neural hidden state for that time point to update the recurrent neural network hidden state; and processing a hidden state of the recurrent neural network at each of one or more time points in the series of time points using an output layer of the recurrent neural network to generate a respective prediction that characterizes the subject. The method.

2. 10. The method of claim 1, wherein for each time point in the series of time points, the time-varying data for that time point characterizes a dosage level of a drug administered to the subject at that time point.

3. For each time point in the series of time points, the time-varying data for that time point are 3. The method of claim 2, wherein the cumulative dosage of the drug administered to the subject is characterized using the

4. 2. The method of claim 1, wherein the series of time points includes a first time point and a second time point, and the time-varying data has a first value at the first time point and a second, different value at the second time point.

5. 10. The method of claim 1, wherein the one or more predictions characterizing the subject include, for each time point in the series of time points, a predicted concentration value of the drug in the subject's plasma at that time point.

6. 6. The method of claim 5, wherein, for each time point in the series of time points, generating a predicted concentration value of the drug in the subject's plasma at that time point comprises processing a hidden state of the recurrent neural network at that time point using an output layer of the recurrent neural network to generate a predicted concentration value of the drug in the subject's plasma at that time point.

7. The method of claim 1, wherein initializing the hidden state of the recurrent neural network based on the MLP output includes setting the hidden state of the recurrent neural network equal to the MLP output.

8. The method described in claim 1, wherein the time-invariant data characterizes the age of the subject.

9. The method described in claim 1, wherein the time-invariant data characterizes the gender of the subject.

10. The method of claim 1 , wherein the multi-layer perceptron comprises multiple hidden layers.

11. 2. The method of claim 1 , wherein, at an end of the series of time points, processing the hidden state of the recurrent neural network at that time point using an output layer of the recurrent neural network to generate a respective prediction that characterizes the subject comprises processing both the hidden state of the recurrent neural network at that time point and the MLP output using the output layer of the recurrent neural network to generate a respective prediction that characterizes the subject.

12. 12. The method of claim 11 , wherein processing both the current hidden state of the recurrent neural network and the MLP output using an output layer of the recurrent neural network to generate respective predictions characterizing the subject comprises summing the current hidden state of the recurrent neural network and the MLP output.

13. The method of claim 1 , wherein the recurrent neural network includes a long short-term memory neural network layer.

14. The method of claim 1 , wherein the recurrent neural network comprises a gated recurrent unit neural network layer.

15. 1. A system comprising: one or more computers; one or more storage devices communicatively coupled to the one or more computers, the one or more storage devices storing instructions that, when executed by the one or more computers, cause the one or more computers to perform the following operations: receiving data characterizing an object; and processing data characterizing the subject to generate one or more predictions characterizing the subject; The data characterizing the subject are: (i) time-varying data characterizing a subject, including respective time-varying data for each time point in a series of time points; (ii) includes time-invariant data characterizing the subject that is the same for each time point in the series of time points; Processing data characterizing the subject includes: processing the time-invariant data using a multi-layer perceptron (MLP) to generate an MLP output; Initializing hidden states of a recurrent neural network based on the MLP outputs; initializing hidden states of a recurrent neural network based on the MLP output, and then processing the time-varying data using the recurrent neural network to generate one or more predictions that characterize the subject; For each time point in the series of time points, processing the time-varying data for the time point and the recurrent neural hidden state to update a hidden state of the recurrent neural network; processing a hidden state of the recurrent neural network at the time point using an output layer of the recurrent neural network to generate, for each of one or more time points of the series of time points, a respective prediction that characterizes the subject; The system.

16. One or more non-transitory computer-readable media storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations: It contains the following instructions: receiving data characterizing an object; and processing data characterizing the subject to generate one or more predictions characterizing the subject; The data characterizing the subject are: (i) time-varying data characterizing a subject, including respective time-varying data for each time point in a series of time points; (ii) includes time-invariant data that is the same for each time point in a series of time points; Processing data characterizing the subject includes: processing the time-invariant data using a multi-layer perceptron (MLP) to generate an MLP output; Initializing hidden states of a recurrent neural network based on the MLP outputs; initializing hidden states of a recurrent neural network based on the MLP output, and then processing the time-varying data using the recurrent neural network to generate one or more predictions that characterize the subject; For each time point in the series of time points, processing the time-varying data for the time point and the recurrent neural hidden state to update a hidden state of the recurrent neural network; processing a hidden state of the recurrent neural network at the time point using an output layer of the recurrent neural network to generate, for each of one or more time points of the series of time points, a respective prediction that characterizes the subject; said one or more non-transitory computer-storage media;

17. 17. The non-transitory computer system of claim 16, wherein for each time point in the series of time points, the time-varying data for that time point characterizes a dosage level of a drug administered to the subject at that time point. Data storage media.

18. 20. The non-transitory computer storage medium of claim 17, wherein for each time point in the series of time points, the time-varying data for that time point characterizes a cumulative dosage of drug administered to the subject at that time point.

19. 17. The non-transitory computer storage medium of claim 16, wherein the series of time points includes a first time point and a second time point, and the time-varying data has a first value at the first time point and a second, different value at the second time point.

20. 17. The non-transitory computer storage medium of claim 16, wherein the one or more predictions characterizing the subject include, for each time point in the series of time points, a predicted concentration value of the drug in the subject's plasma at that time point.

Images