Abstract: A reinforcement learning (RL) based adaptive state observation model usable for implementing a brain machine interface (BMI) is proposed for decoding a brain signal to determine a movement action and controlling a machine to perform the movement action. In the model, the brain signal is processed by a neural network (NN) for applying a nonlinear mapping defined by NN weights to the brain signal to thereby yield a transformed brain signal. The NN learns the nonlinear mapping by RL, allowing the weights to be adaptively and continuously updated to follow nonlinearity and non-stationarity of the brain signal. The transformed brain signal is processed by a Kalman filter (KF) to yield a control signal for controlling the machine to perform the movement action, thereby utilizing the KF to provide smooth generation of the control signal while blocking adverse influence of nonlinearity and non-stationarity of the brain signal to the KF.

Abstract: An explanation of a detection/classification algorithm made using a deep learning neural network clarifies the results that are formed and helps a user to identify the root cause of defect detection/classification model performance issues. A relevance map is determined based on a layer-wise relevance propagation algorithm. A mean intersection over union score between the relevance map and a ground truth is determined. A part of one of the semiconductor images that contributed to the classification using the deep learning model based on the relevance map and the mean intersection over union score is determined.

Abstract: An encoder for encoding weight parameters of a neural network is configured to obtain a plurality of weight parameters of the neural network, to encode the weight parameters of the neural network using a context-dependent arithmetic coding, to select a context for an encoding of a weight parameter, or for an encoding of a syntax element of a number representation of the weight parameter, in dependence on one or more previously encoded weight parameters and/or in dependence on one or more previously encoded syntax elements of a number representation of one or more weight parameters, and to encode the weight parameter, or a syntax element of the weight parameter, using the selected context. Corresponding decoder, quantizer, methods and computer programs are also described.

Abstract: One example method includes performing the various operations concerning a model that is operable to predict resource usage and execution time of computing workloads. The operations include extracting a fingerprint associated with telemetry data, and the telemetry data was generated based on performance of one of the computing workloads, in a constrained infrastructure, checking a fingerprint catalog to determine if there is a same or similar fingerprint to the fingerprint, when the same or similar fingerprint is found in the fingerprint catalog, inferring that the model includes information about the computing workload and the model is able to predict telemetry data and execution time for the computing workload in a target infrastructure, and when the same or similar fingerprint is not found, inserting the extracted fingerprint into the fingerprint catalog, and generating a retrained model by retraining the model using the telemetry data associated with the extracted fingerprint.

Abstract: A neural network processor is configured to execute instructions in parallel on different computing engines (CEs) to perform convolution operations on an input dataset. The input dataset is divided into overlapping chunks, including a first chunk and a second chunk. Each CE processes a last portion of a respective chunk to compute respective shared states and receives additional shared states for processing. The first chunk is the respective chunk for a first CE. The second chunk is the respective chunk for a second CE. The additional shared states received by the second CE are the respective shared states computed by the first CE. The second CE receives the respective shared states computed by the first CE as a substitute for intermediate states that would otherwise have been computed by the second CE using a first portion of the second chunk.

Abstract: A method includes receiving a source data set and a target data set and identifying a loss function for a deep learning model based on the source data set and the target data set. The loss function includes encoder weights, source classifier layer weights, target classifier layer weights, coefficients, and a policy weight. During a first phase of each of a plurality of learning iterations for a learning to transfer learn (L2TL) architecture, the method also includes: applying gradient decent-based optimization to learn the encoder weights, the source classifier layer weights, and the target classifier weights that minimize the loss function; and determining the coefficients by sampling actions of a policy model. During a second phase of each of the plurality of learning iterations, determining the policy weight that maximizes an evaluation metric.

Abstract: A model construction support system supports searching for a feature used to construct a prediction model that outputs an objective variable related to a predicted event for a machine based on explanatory variables, and a division method for dividing the explanatory variables into groups to improve calculation accuracy of the objective variables based on the prediction model. The system divides the explanatory variables into a plurality of groups, calculates accuracy of the features set based on the explanatory variable in the groups, and calculates a score of the feature in the groups based on the accuracy and a support ratio of the explanatory variable to all of the explanatory variables before division. The system calculates accuracy of a group division feature used to divide the explanatory variables, and a score in the groups based on the score and the accuracy in the groups.

Abstract: Methods, systems, and computer programs are presented for determining parameters of neural networks and selecting embedding dimensions for the feature fields. One method includes an operation for initializing parameters of a neural network and weights for embedding sizes for each feature associated with the neural network. The parameters of the neural network and the weights are iteratively optimized. Each optimization iteration comprises training the neural network with current parameters of the neural network to optimize a value of the weights, and training the neural network with current values of the weights to optimize the parameters of the neural network. Further, the method includes operations for selecting embedding sizes for the features based on the optimized values of the weights, and for training the neural network based on the selected embedding sizes for the features to obtain an estimator model. A prediction is generated utilizing the estimator model.

Abstract: Performing a goal-seek analysis of spatial-temporal data by generating a hierarchical cluster according to spatial temporal data, determining a spatial-temporal location input for a target, determining spatial-temporal predictor values for the spatial-temporal location, and adjusting the hierarchical cluster according to and the spatial-temporal predictors.

Abstract: Aspects and examples disclosed herein are directed to data enrichment on insulated appliances. An appliance for performing knowledge mining in a disconnected state is operative to: import a plurality of containerized cognitive functions and a seed index from a service node, when the appliance is initially connected to a network; ingest data of a first type from a first data source coupled to the appliance; enrich at least a first portion of the ingested data, when in the disconnected state, with at least one of the cognitive functions of the plurality of containerized cognitive functions; identify at least a second portion of the ingested data for enrichment by a cognitive function that is not within the plurality of containerized cognitive functions; based at least on knowledge extracted from the enriched data, further enrich the enriched data; and grow the seed index into an enhanced index with the enriched data.

Abstract: Systems and methods for user-specific content generation can leverage parameter tuning based on user feedback data to tune a set of parameters for conditioning a machine-learned content generation model for the content generation. The set of parameters can be processed with the machine-learned content generation model to generate a model-generated content item that is associated with user tastes and interests. The parameter tuning can include processing user interactions associated with the model-generated content item to adjust the set of parameters.

Abstract: A method for training a machine learning model for determining a quality grade of data sets from each of a plurality of sensors. The sensors are configured to generate surroundings representations. The method includes: providing data sets of each of the sensors from corresponding surroundings representations; providing attribute data of ground truth objects of the surroundings representations; determining a quality grade of the respective data set of each of the sensors using a metric, the metric comparing at least one variable, which is determined using the respective data set, with at least one attribute datum of at least one associated ground truth object of the surroundings representation; and training the machine learning model using the data sets of each of the sensors and the respectively assigned determined quality grades.

Abstract: Systems, devices, and methods related to a Deep Learning Accelerator and memory are described. For example, an integrated circuit device may be configured to execute instructions with matrix operands and configured with random access memory (RAM). A compiler has an artificial neural network configured to identify an optimized compilation option for an artificial neural network to be compiled by the compiler and/or for a hardware platform of Deep Learning Accelerators. The artificial neural network of the compiler can be trained via machine learning to identify the optimized compilation option based on the features of the artificial neural network to be compiled and/or features of the hardware platform on which the compiler output will be executed.

Abstract: An electronic apparatus includes at least one memory configured to store at least one instruction and a first neural network model; a communicator comprising communication circuitry; and at least one processor configured to execute the at least one instruction to: receive, from an external electronic device, information on a second neural network model stored in the external electronic device through the communicator; compare the first neural network model with the second neural network model based on the information on the second neural network model; and control the communicator to transmit, to the external electronic device, information on a weight between nodes of the first neural network model based on comparison between the second neural network model and the first neural network model.

Abstract: Embodiments provide a machine learning model to identify changes in setpoints of one or more environmental control modules that, while having a high critical error, provide greater reductions in a cost function associated with the one or more environmental modules provided in an environmentally controlled space. The high critical error associated with the identified changes may still be within an acceptable threshold range associated with the environmentally controlled space. Thus, contrary to rule-based methods, the artificial intelligence (AI) based model described herein may recommended optimal changes to the system that yield to greater savings in the cost function and may not focus on minimizing the critical control error. Rather, the AI-based technique may simply aim at keeping the critical control error within an acceptable threshold range.

Abstract: Provided is an apparatus for training a knowledge tracking model, which is an apparatus for predicting a correct answer probability of a user on the basis of data augmentation, the apparatus including: a problem-solving data storage unit configured to store problem-solving data in which a problem solved by a user and a response of the user to the problem are mapped; a data augmentation performing unit configured to receive the problem-solving data from the problem-solving storage unit and convert the problem-solving data to generate augmented data; a regularization performing unit configured to receive the augmented data from the data augmentation performing unit and perform a regularization operation using a regularization loss function determined on the basis of a data augmentation method that is performed; and a model training unit configured to input the augmented data to a knowledge tracking model, allow the knowledge tracking model to learn a weight representing a relationship between a problem-solving

Abstract: Techniques are disclosed for providing a scalable multi-tenant serve pool for chatbot systems. A query serving system (QSS) receives a request to serve a query for a new skillbot. The QSS comprises a plurality of deployments, each of which is configured to host a plurality of machine-learning models, each machine-learning model being associated with a skillbot, each deployment including a serving container and a model manager container that hosts a model manager, the serving container including a plurality of sub-containers, each of which hosts one of the machine-learning models downloaded by the model manager. The QSS selects a first deployment to be assigned to the new skillbot based on a first criterion, and loads the machine-learning model associated with the new skillbot into the first deployment. The machine-learning model is trained to serve the query for the new skillbot. The query is served using the machine-learning model.

Abstract: Systems and methods for natural language processing are described. Embodiments are configured to receive a structured representation of a search query, wherein the structured representation comprises a plurality of nodes and at least one edge connecting two of the nodes, receive a modification expression for the search query, wherein the modification expression comprises a natural language expression, generate a modified structured representation based on the structured representation and the modification expression using a neural network configured to combine structured representation features and natural language expression features, and perform a search based on the modified structured representation.

Abstract: Aspects and examples disclosed herein are directed to data enrichment on insulated appliances. An appliance for performing knowledge mining in a disconnected state is operative to: ingest data of a first type from a first data source coupled to the appliance; enrich at least a first portion of the ingested data, when in the disconnected state, with at least one of the cognitive functions of the plurality of containerized cognitive functions; identify at least a second portion of the ingested data for enrichment by a cognitive function that is not within the plurality of containerized cognitive functions; store the enriched data in an index; triage the ingested data and the enriched data in the index for uploading; upload the triaged data when reconnected to a network; and import an updated plurality of containerized cognitive functions when reconnected.

Abstract: Disclosed herein are embodiments of systems, methods, and products comprises a server for efficiently processing electronic requests. The server receives a plurality of predictive computer models and generates a specification file for each model by parsing the source code of each model. When the server receives an electronic request, the API layer of the server validates the request by verifying the inputs of the request satisfying validation codes in the specification file of the corresponding model. If the electronic request is invalid, the server imputes valid values for the request and sends the imputed values to the model execution layer. Within the model execution layer, the server utilizes an integrated development environment of a third-party server to call the function of the corresponding model. The model execution layer transmits the function's output results back to the API layer, which transmits the output results to the user device.