Abstract: A reader apparatus detects a communication object by radiating radio waves and receiving a response, determines a change in a state based on at least one of a position and an inclination of the reader apparatus, and controls power consumed in the reader apparatus by setting a length of a first period, in which a reduction of power consumed in the detection is not performed, and a length of a second period, in which a reduction of the power is performed, based on the change in the state.

Abstract: Embodiments disclosed herein generally relate to line-powered wireless communications systems, and more specifically to methods and apparatus for providing persistent and ubiquitous wireless communications and sensor networks in physical premises to enable a wide variety of different applications and use cases.

Abstract: This disclosure relates to systems with multiple NFC front ends and at least one shared controller. One disclosed system includes a first discrete device having a first near field communication (NFC) controller, a second discrete device having a second NFC controller, a communicative connection between the first discrete device and the second discrete device, and a system controller located on the first discrete device. The system controller is communicatively connected to the first NFC controller. The system controller is connected to the second NFC controller via the communicative connection. In specific approaches, the second discrete device can include a display which is also controlled by the system controller, and the system controller is NCI compliant and application aware. In specific approaches, the system controller can control the second NFC controller based on the content concurrently being provided to the display.

Abstract: A communication apparatus includes an antenna, an actuator assembly, a transducer, and a controller. The actuator assembly is configured to move the antenna in two dimensions relative to a space configured to contain a wireless tag. The transducer is configured to provide phase data indicating a phase of a signal transmitted from the wireless tag in at least (a) a first detection condition in which the antenna has a first position relative to the space and (b) a second detection condition in which the antenna has a second position relative to the space. The controller is configured to determine whether the wireless tag is located inside of the space based on the phase data.

Abstract: A method and a card reader for extracting Card Holder Data, CHD, from an Integrated Circuit Card, ICC, on a payment card. The card reader sends a request for a file list of Application Identifier's, AID's, stored in the ICC of the payment card, and checks if a Short File Identifier, SFI, associated with the AID is stored in a memory of the card reader. If not stored a Get Processing Options, GPO, command is executed to extract the SFI from the ICC, which is stored in the memory. Then the card reader 6 reads the CHD files according to the SFI and extracts a Primary Account Number, PAN. If CHD is already stored the card reader reads the CHD files according to SFI and extracts PAN directly.

Abstract: This invention is for tracking at least one plant. A method of this invention comprises: putting at least one seed or at least one stem in a corresponding at least one pot; positioning a corresponding at least one RFID tag with respect to the seed or stem in a manner, wherein the RFID tag comprises a strap; packaging a harvested material into a packaged product and attaching the RFID tag from the potted plant, or a product RFID tag that is associated with the plurality of tags to the packaged product; confirming a request for authorization by a RFID buy card; and after confirming ID information, transferring at least one product.

Abstract: Disclosed are a reader device, system, and method for communicating with a wireless sensor. The reader device may be configured to analyze the strength of a response signal transmitted from the wireless sensor in response to an excitation pulse generated by the reader device. In one embodiment, the reader device may be configured to engage be placed in a plurality of modes to allow the reader to transmit a signal, such as a short pulse of energy or a short burst of radio frequency energy to cause the wireless sensor to output a resonant signal. The reader device may receive the resonant signal from the wireless sensor and evaluate it against predetermined values. The evaluated signals may be used to assess the strength and the proximity of the reader device relative to the wireless sensor as it is implanted in a patient.

Abstract: Systems and methods for tracking objects in space are disclosed. The systems and methods include capturing two-dimensional (2D) image data from which a barcode is decoded and capturing, generating, or otherwise accessing three-dimensional (3D) image data from which a 3D object is identified. A 2D image of a barcode and barcode data is combined with the 3D object to form reference 3D object data that is used for comparison to subsequently captured 3D and 2D image data. In some examples, a four-dimensional (4D) projection of the reference 3D object data is used for comparison and validation of the subsequently captured 3D and 2D image data.

Abstract: Embodiments of the present disclosure include near co-axial polarized illuminator apparatuses and uses thereof. The near co-axial polarized illuminator is internally positionable in near co-axial alignment with one or more imager(s) utilized to capture a representation of a field of view illuminated by the near co-axial polarized illuminator. Embodiments include small form factor reader(s) that utilize a near co-axial polarized illuminator to improve illumination of a field of view captured via multiple imagers for detecting and/or decoding machine-readable symbologies, such as barcodes. Some embodiments include a DPM channel image sensor, a standard range channel image sensor, a near co-axial polarizer light source positioned adjacent to the DPM channel image sensor and adjacent to the standard range channel image sensor, a near co-axial polarizer aligned with the near co-axial polarizer light source, and an analyzer aligned with the DPM channel image sensor and/or the standard range channel image sensor.

Abstract: Embodiments position a barcode reader, or other optical detector, to scan upright barcodes when the user's hand is in the neutral position. To do this, embodiments rest the body of the ring scanner device on top of the index finger, with the palm facing inward toward the user's body. Embodiments have integrated interactivity, including a touchscreen display and separate buttons that can be used to operate the optical detector and input data into an integrated computing device. The integrated computing device can run software applications on board the ring scanner device. Finally, the integrated computing device can be remotely managed.

Abstract: Methods and apparatus for using an indicator window of a handheld scanner as a trigger are disclosed herein. An example handheld scanner includes: a housing; an image sensor to capture image data through a front-facing opening; an indicia decoder; an indicator window positioned to face generally away from the front-facing opening and toward a user when the handheld scanner is in a handheld position; a light source disposed inside the housing to emit indication light through the window to provide an indication; a light detector disposed inside the housing and positioned to detect a reflection of the indication light received from an object positioned in front of or on the window outside the housing; and a processor configured to control a mode of the handheld scanner and/or a device in communication with the handheld scanner in response to the light detector detecting the reflection of the emitted indication light.

Abstract: A code associated with a managed device is scanned by a mobile device operated by service personnel. The code at least comprises an identifier for managed device. A current geographical location of the mobile device and the code are provided to a cloud/server. The location is associated with the managed device and updated or added to a shared private blockchain ledger. The shared private blockchain ledger shared between one or more enterprises and one or more service organizations. In an embodiment, the managed device dynamically generates the code and presents the code on a display associated with the managed device for scanning by the mobile device.

Abstract: One example of the present disclosure can include a server that can determine transport routes for individual objects based on rules. The transport routes can include an intermediary at which the individual objects are to be consolidated into a single container. The server can also receiving a request from a client computer for an object label corresponding to one of the individual objects. In response to receiving the request, the server can determine the intermediary based on a transport route for the object, generate the object label based on the intermediary, and transmit the object label to the client computer for causing the object label to be printed with a printer and coupled to the object.

Abstract: Methods, systems, and apparatus for combining preprinted information together with coded sensor information within a two-dimensional barcode. The sensor information may be of an environmental, physical or biological nature, and records a cumulative change in status of the environmental or biological condition to which the labeled product has been exposed. A sensor dye chemistry is employed that undergoes a continuous chemical or physical state change in response to the occurrence of the environmental condition. The continuous change is between an initial state and an end state causing a change in the color state of the sensor dye embedded within the sensor-augmented two-dimensional barcode, encoding sensor digital information. Sensor information is recovered utilizing the error-correction feature during barcode decoding.

Abstract: A system for entanglement-enhanced machine learning with quantum data acquisition includes a first variational circuit that generates a plurality of entangled probe light fields that interacts with a sample and is then processed by a second variational quantum circuit to produce at least one detection light field, a detector is used to measure a property of the at least one detection light field, and the first and second variational quantum circuits are optimized though machine learning. A method for entanglement-enhanced machine learning with quantum data acquisition includes optimizing a setting of a first and second variational quantum circuits, which includes probing a training-set with a plurality of entangled probe light fields generated by the first variational quantum circuit, and measuring a phase property of at least one detection light fields generated by the second variational quantum circuit from the plurality of entangled probe light fields after interaction with the training-set.

Abstract: Methods, systems and apparatus for implementing iSWAP quantum logic gates between a first qubit and a second qubit. In one aspect, a method includes implementing a cascade schedule that defines a trajectory of a detuning between a frequency of the first qubit and a frequency of the second qubit. Implementing the cascade schedule includes: during a first stage, adiabatically driving detuning between the frequency of the first qubit and the frequency of the second qubit through a first avoided crossing in a leakage channel; during a second stage, driving detuning between the frequency of the first qubit and the frequency of the second qubit through a second avoided crossing in a swap channel; during a third stage, evolving the first qubit and second qubit; during a fourth stage, implementing the second stage in reverse order; and during a fifth stage, implementing the first stage in reverse order.

Abstract: A system and method for initializing and optimizing a variational quantum circuit on a hybrid quantum-classical computer, comprising a set of gates and a set of initial parameters representing a model of a physical system. A quantum circuit is generated comprising a set of smaller contiguous subcomponents which can be independently optimized to minimize a property of the physical system, such as ground state energy or the absorption spectrum of a molecule. At least one entangling gate is introduced between at least two circuit subcomponents. The initial parameters of the circuit components may be set according to values obtained from a parameter library. Once the initial parameters are set, the circuit components of the quantum computer proceed to optimization, which is independent for each subcomponent of the system. The optimization method may also include the use of a variational quantum eigensolver (VQE).

Abstract: A quantum circuit generation method includes determining a reference state of a target molecule and N excitations states corresponding to the reference state, where N is a positive integer greater than or equal to 1; determining M excitations states from the N excitations states based on an attribute of the reference state and attributes of the N excitations states, where M is a positive integer greater than or equal to 1 and less than or equal to N; and generating a first quantum circuit based on the M excitations states such as to improve computation efficiency and to reduce resource consumption.

Abstract: There is described a method for performing a two-qubit gate. The method comprises coupling a first qubit to a second qubit, and coupling a third qubit to the first qubit and to the second qubit, wherein the third qubit is a coupler. A drive signal is applied to the coupler to perform the two-qubit gate on the first qubit and the second qubit by driving the coupler, the drive signal creating coupling between eigenstates of a Hamiltonian of the coupler.

Abstract: A method of preparing an entangled state of a plurality of qubits. The method comprises: (i) preparing an array of qubits, each qubit being in a ground qubit state and having an excited qubit state.

Abstract: One or more computing devices, systems, and/or methods for a model training framework are provided. A definition and configuration of a model are received. Computations that the model will perform during training of the model are setup based upon the definition and configuration. Summary statistics to be tracked during training are specified. A batch of training data is input into the model to train the model based upon hyper parameters specified in the configuration of the model, and the summary statistics are tacked during the training. The parameters of the model are updated based upon a function corresponding to accuracy of the model processing the training data. The summary statistics are outputted.

Abstract: A machine learning platform operating at a server is described. The machine learning platform accesses a dataset from a datastore. A task that identifies a target of a machine learning algorithm from the machine learning platform is defined. The machine learning algorithm forms a machine learning model based on the dataset and the task. The machine learning platform deploys the machine learning model and monitors a performance of the machine learning model after deployment. The machine learning platform updates the machine learning model based on the monitoring.

Abstract: The invention relates to a method of performing a process using artificial intelligence. The method comprises running, by a computing device, an application configured to perform a process which uses an artificial intelligence model for processing signals and defining at least one parameter set for performing the at least one process; running, by the computing device, a user-driven workflow engine which comprises multiple modules including at least a first and second module; defining, by the first module, a context of the process and generating corresponding context information, providing the artificial intelligence model based on the generated context information of the process; and using, by the second module, the artificial intelligence model in a user-driven workflow within the application while executing the process. Advantageously, the combination of these modules describes an end-to-end connection which is self-learning and self-improving and is targeted at users with no expertise in the AI domain.

Abstract: A machine learning network is implemented by executing a computer program of instructions on a machine learning accelerator (MLA) comprising a plurality of interconnected storage elements (SEs) and processing elements (PEs. The instructions are partitioned into blocks, which are retrieved from off-chip memory. The block includes a set of deterministic instructions to be executed by on-chip storage elements and/or processing elements according to a static schedule. The block also includes the number of non-deterministic instructions to be executed prior to executing the set of deterministic instructions in this block. These non-deterministic instructions may be instructions for storage elements to retrieve data from off-chip memory and are contained in one or more prior blocks. The execution of these non-deterministic instructions is counted, for example through the use of tokens.

Abstract: A hyperscale artificial intelligence and machine learning infrastructure includes a plurality of racks, where: at least one or more of the racks include one or more GPU servers; at least one or more of the racks include one or more storage systems; each of the racks include one or more switches coupled to at least one switch in another rack; and the one or more GPU servers are configured to execute one or more artificial intelligence or machine learning applications, wherein data stored within the one or more storage systems is used as input to the one or more artificial intelligence or machine learning applications.

Abstract: The techniques herein include using an input context to determine a suggested action. One or more explanations may also be determined and returned along with the suggested action. The one or more explanations may include (i) one or more most similar cases to the suggested case (e.g., the case associated with the suggested action) and, optionally, a conviction score for each nearby cases; (ii) action probabilities, (iii) excluding cases and distances, (iv) archetype and/or counterfactual cases for the suggested action; (v) feature residuals; (vi) regional model complexity; (vii) fractional dimensionality; (viii) prediction conviction; (ix) feature prediction contribution; (x) conviction ratio; (xi) contribution ratio; and/or other measures such as the ones discussed herein, including certainty. In some embodiments, the explanation data may be used to determine whether to perform a suggested action.

Abstract: A learning model management system executes a provisional evaluation when the number of feedback data is equal to or less than a threshold of a definite evaluation but is more than a threshold of a provisional evaluation. In this provisional evaluation, prediction accuracy of the learning model that is in operation is evaluated a plurality of times, and whether or not the prediction accuracy of the learning model that is in operation is in a deterioration trend is determined based on the change tendency of the evaluation results. If it is determined that the prediction accuracy of the learning model that is in operation is in a deterioration trend, the learning model management system provides a notification about the deterioration trend of the prediction accuracy to a manager to cause him/her to increase the number of feedback data up to a number that enables the definite evaluation.

Abstract: A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform: receiving historical delivery records over a predetermined time period from partners associated with items offered to subregions through an online platform; generating nodes for combinations each comprising a respective one of the partners, a respective one of the items offered by the partners, and a respective one of the subregions; generating, using a machine learning model, a respective classification for each respective node on whether to tag the each respective node as deliverable in a predetermined time window; and automatically tagging a portion of the nodes as deliverable in the predetermined time window in the online platform. Other embodiments are disclosed.

Abstract: A system is described that determines, based on information associated with a user of a computing device, an event for initiating an interaction between the user and an assistant executing at the computing device. The system selects, based on the event and from a plurality of actions performed by the assistant, at least one action associated with the event. The system determines, based on the at least one action, whether to output a notification of the event which includes an indication of the event and a request to perform the at least one action associated with the event. Responsive to determining to output the notification of the event, the system sends, to the assistant, the notification of the event for output during the interaction between the user and the assistant.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for searching an autonomous vehicle sensor data repository. One of the methods includes maintaining a collection of sensor samples and one or more embeddings of each sensor sample. Each sensor sample is generated from sensor data at multiple time steps and characterizes an environment at each of the multiple time steps. Each embedding corresponds to a respective portion of the sensor sample and has been generated by an embedding neural network. A query specifying a query portion of a query sensor sample is received. A query embedding corresponding to the query portion of the query sensor sample is generated through the embedding neural network. A plurality of relevant sensor samples that have embeddings that are closest to the query embedding are identified as characterizing similar scenarios to the query portion of the query sensor sample.

Abstract: Defect resistant designs for location-sensitive neural network processor arrays are provided. In various embodiments, plurality of neural network processor cores are arrayed in a grid. The grid has a plurality of rows and a plurality of columns. A network interconnects at least those of the plurality of neural network processor cores that are adjacent within the grid. The network is adapted to bypass a defective core of the plurality of neural network processor cores by providing a connection between two non-adjacent rows or columns of the grid, and transparently routing messages between the two non-adjacent rows or columns, past the defective core.

Abstract: A method implemented by one or more computing systems includes accessing a first data matrix including a plurality of row data and a plurality of column data. The method further includes providing, to a first generative adversarial network (GAN), a first data input including a plurality of row vectors corresponding to the plurality of row data, and providing, to a second GAN, a second data input including a plurality of column vectors corresponding to the plurality of column data. The method further includes generating, by simultaneous co-clustering the plurality of row vectors and the plurality of column vectors by the first GAN and the second GAN, a co-clustered correlation matrix based on the plurality of row vectors and the plurality of column vectors. The method further includes the co-clustered correlation matrix includes co-clustered associations between the plurality of row data and the plurality of column data.

Abstract: Systems and/or techniques for facilitating image forgery detection via headpose estimation may include a system that can receive a document from a client device. The system can identify, by executing a first trained machine learning model, an object that is depicted in the document. The system can determine, by executing a second trained machine learning model, a pose of the object. The system can determine, by executing a third trained machine learning model, whether the document is authentic or forged based on the pose of the object. The system can, in response to determining that the document is forged, transmit an unsuccessful validation message to the client device.

Abstract: A system includes a neural network that includes a Mixture of Experts (MoE) subnetwork between a first neural network layer and a second neural network layer. The MoE subnetwork includes multiple expert neural networks. Each expert neural network is configured to process a first layer output generated by the first neural network layer to generate a respective expert output. The MoE subnetwork further includes a gating subsystem that selects, based on the first layer output, one or more of the expert neural networks and determine a respective weight for each selected expert neural network, provides the first layer output as input to each of the selected expert neural networks, combines the expert outputs generated by the selected expert neural networks in accordance with the weights for the selected expert neural networks to generate an MoE output, and provides the MoE output as input to the second neural network layer.

Abstract: Methods and apparatuses for implementing a neural network using symmetric tensors. In embodiments, a system may include a higher order neural network with a plurality of layers that includes an input layer, one or more hidden layers, and an output layer. Each of the input layer, the one or more hidden layers, and the output layer includes a plurality of neurons, where the plurality of neurons includes at least first order neurons and second order neurons, and where inputs at a second order neuron are combined using a symmetric tensor.

Abstract: A phase change material (PCM)-based neural network device according to an embodiment comprises: a plurality of neurons disposed for input layers and output layers, respectively; a plurality of PCMs connecting input lines of the input layers and output lines of the output layers; and at least one backward spike generator (BSG) shared by the plurality of neurons, and generating spike on the basis of an output pulse outputted from each of the neurons of the output layers.

Abstract: Systems and methods for heterogenous hardware acceleration are disclosed. The systems and methods can include a neural network processing unit comprising compute tiles. Each of a first set of the compute tiles can include a first tensor array configured to support operations in a first number format. Each of a second set of the compute tiles can include a second tensor array configured to support operations in a second number format, the second number format supporting a greater range or a greater precision than the first number format, and a de-quantizer configured to convert data in the first number format to data in the second number format. The systems and methods can include neural network processing units, multi-chip hardware accelerators and distributed hardware accelerators including low-precision components for performing interference tasks and high-precision components for performing training tasks.

Abstract: A signal processing method and apparatus, where the apparatus includes an input interface configured to receive an input signal matrix and a weight matrix, a processor configured to interleave the input signal matrix to obtain an interleaved signal matrix, partition the interleaved signal matrix, interleave the weight matrix to obtain an interleaved weight matrix, process the interleaved weight matrix to obtain a plurality of sparsified partitioned weight matrices, perform matrix multiplication on the sparsified partitioned weight matrices and a plurality of partitioned signal matrices to obtain a plurality of matrix multiplication results, and an output interface configured to output a signal processing result.

Abstract: Array-integrated upstream/downstream routers for circuit-switched parallel connectivity are provided. A system comprises an array of neural cores having at least one dimension, a plurality of signal wires, and a plurality of routers. Each neural core comprises a plurality of ordered input wires, a plurality of ordered output wires, and a plurality of synapses, each synapse operatively coupled to one of the plurality of input wires and one of the plurality of output wires. The plurality of signal wires are disposed along each dimension of the array of neural cores. Each router is operatively coupled to one of the plurality of neural cores and to at least one signal wire along each dimension of the array of neural cores. Each of the plurality of routers is adapted to selectively route a signal from the at least one signal wire to its coupled neural core. Each of the plurality of routers is adapted to selectively route a signal from its coupled neural core to the at least one signal wire.

Abstract: Techniques for input adaptation from disparate data sources for heterogeneous machine learning model execution are described. A preprocessing adapter can perform preprocessing of data obtained from edge devices to suit the input data characteristic requirements of one or more machine learning (ML) models. The preprocessing adapter can determine the input data characteristic requirements in a variety of ways, such as via analysis of the input layer of a ML model or through data variation testing and associated feedback resulting from output data generated by the ML model.

Abstract: Embodiments of the present invention provide a machine learning model training method, including: obtaining target task training data and N categories of support task training data; inputting the target task training data and the N categories of support task training data into a memory model to obtain target task training feature data and N categories of support task training feature data; training the target task model based on the target task training feature data and obtaining a first loss of the target task model, and separately training respectively corresponding support task models based on the N categories of support task training feature data and obtaining respective second losses of the N support task models; and updating the memory model, the target task model, and the N support task models based on the first loss and the respective second losses of the N support task models.

Abstract: An example method of training a neural network includes defining hardware building blocks (HBBs), neuron equivalents (NEQs), and conversion procedures from NEQs to HBBs; defining the neural network using the NEQs in a machine learning framework; training the neural network on a training platform; and converting the neural network as trained into a netlist of HBBs using the conversion procedures to convert the NEQs in the neural network to the HBBs of the netlist.

Abstract: Methods, systems, and non-transitory computer readable medium are provided for long short-term memory (LSTM) anomaly detection for multi-sensor equipment monitoring. A method includes training a LSTM recurrent neural network (RNN) model for semiconductor processing fault detection. The training includes generating training data for the LSTM RNN model and providing the training data to train the LSTM RNN model on first training input and first target output to generate a trained LSTM RNN model for the semiconductor processing fault detection. The training data includes the first training input and the first target output based on normal runs of manufacturing processes of semiconductor processing equipment. Another method includes providing input based on runs of manufacturing processes of semiconductor processing equipment to a trained LSTM RNN model; obtaining one or more outputs from the trained LSTM RNN model; and using the one or more outputs for semiconductor processing fault detection.

Abstract: A system for fault diagnosis is provided. The system may acquire a vibration signal of a target device, and determine one or more feature values of the vibration signal. The system may further determine a fault condition of the target device by applying a fault diagnosis model to the feature values. The fault diagnosis model may include a trained first component including a plurality of stacked trained RBMs, and a trained second component connected to the trained first component. The trained second component may include a trained fully connected layer and a trained output layer connected to the trained fully connected layer.

Abstract: The present application describes a method for training a neural network via deep reinforcement learning (DRL) in a RF network. The method includes a step of receiving, via the neural network, a policy from a third party. The method also includes a step of receiving, via the neural network, features of plural telecommunication groups located in an RF network. The method also includes a step of observing, via the neural network, a graphical representation of the received features of the plural telecommunication groups in the RF network. The method further includes a step of assigning, based on the observation, one of the plural telecommunication groups to one of plural channels in the RF network. The method even further includes a step of determining, via the neural network, a change in throughput of the RF network based on the assignment. The method yet even further incudes as step of adjusting, based on the determined change in throughput, the policy received from the third party.

Abstract: An object is to obtain a trained model whose robustness is high in a case of learning a network in an image signal processing system. For teacher images consisting an acquired teacher image group, a distribution characteristic of at least one of three attributes of color is analyzed and based on analysis results, a teacher image group uniform in the distribution characteristic is generated. Learning of a network is performed by generating a data set including a set of a teacher image and a pupil image from a new teacher image group thus obtained.

Abstract: In a network discovery and management system, a machine learning (ML) DLAD processor trains, validates, updates, and stores machine learning models. A ML training data preparation program performs operations to process and format input data to generate ML training data that can be used to train ML models. ML training program uses the ML training data to train ML models, thereby generating trained ML models. The ML training program can re-train or update the training of ML models as the system collects additional data and produces additional estimates, predictions, and forecasts. ML model validation program performs validation testing on trained ML models to generate one or more metrics that can indicate accuracy of predictions generated by the trained models. The resulting ML model(s) can be used to manage the network including but not limited to retrieve, instantiate and execute dynamic applications based on predictions made based on the models.

Abstract: Examples of the present disclosure provide systems and methods relating to a machine comprehension test with a learning-based approach, harnessing neural networks arranged in a parallel hierarchy. This parallel hierarchy enables the model to compare the passage, question, and answer from a variety of perspectives, as opposed to using a manually designed set of features. Perspectives may range from the word level to sentence fragments to sequences of sentences, and networks operate on word-embedding representations of text. A training methodology for small data is also provided.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a policy neural network having a plurality of policy parameters and used to select actions to be performed by an agent to control the agent to perform a particular task while interacting with one or more other agents in an environment. In one aspect, the method includes: maintaining data specifying a pool of candidate action selection policies; maintaining data specifying respective matchmaking policy; and training the policy neural network using a reinforcement learning technique to update the policy parameters. The policy parameters define policies to be used in controlling the agent to perform the particular task.

Abstract: Disclosed herein are techniques for performing multi-layer neural network processing for multiple contexts. In one embodiment, a computing engine is set in a first configuration to implement a second layer of a neural network and to process first data related to a first context to generate first context second layer output. The computing engine can be switched from the first configuration to a second configuration to implement a first layer of the neural network. The computing engine can be used to process second data related to a second context to generate second context first layer output. The computing engine can be set to a third configuration to implement a third layer of the neural network to process the first context second layer output and the second context first layer output to generate a first processing result of the first context and a second processing result of the second context.