Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting an action to be performed by a reinforcement learning agent. The method includes obtaining an observation characterizing a current state of an environment. For each layer parameter of each noisy layer of a neural network, a respective noise value is determined. For each layer parameter of each noisy layer, a noisy current value for the layer parameter is determined from a current value of the layer parameter, a current value of a corresponding noise parameter, and the noise value. A network input including the observation is processed using the neural network in accordance with the noisy current values to generate a network output for the network input. An action is selected from a set of possible actions to be performed by the agent in response to the observation using the network output.

Abstract: An inference device comprises a weight storage part that stores weights, an input data storage part that stores input data, and a PE (Processing Element) that executes convolution computation in convolutional neural network using the weights and input data. The PE adds up weight elements to be multiplied with elements of the input data for each of variable values of the elements of the input data. The PE multiplies each of the variable values of the elements of the input data with each cumulative sum value of weights corresponding to the variable values of the elements of the input data. The PE adds up a plurality of multiplication results obtained by the multiplications.

Abstract: A method of constructing an adaptive multiply accumulate layer in a convolutional neural network, including determining an activation data map width, an activation data map height, a channel depth, a batch, a kernel width, a kernel height and a filter set number, setting a first dimension of an adaptive multiplier layer based on the activation data map width, setting a second dimension of the adaptive multiplier layer based on the channel depth, setting a third dimension of the adaptive multiplier layer based on the filter set number and constructing the adaptive multiplier layer based on the first dimension, the second dimension and the third dimension.

Abstract: Disclosed are a data flow method and apparatus for neural network computation. The data flow method for neural network computation includes initializing the lifecycle of a variable in a computational graph; and defining a propagation rule for a variable in use to flow through a node. A definition of the variable is produced at a precursor node of the node, such that an input set of valid variables flowing through the node contains the variable. The method may be used on neural network computation in a deep learning training system.

Abstract: An apparatus to facilitate optimization of a convolutional neural network (CNN) is disclosed. The apparatus includes optimization logic to receive a CNN model having a list of instructions and including pruning logic to optimize the list of instructions by eliminating branches in the list of instructions that comprise a weight value of 0.

Abstract: Methods and systems for a reinforcement learning system. A spatial and temporal representation of an observed state of an environment is encoded. A previous state is estimated from a given state and a size of a reward is adjusted based on a difference between the estimated previous state and the previous state.

Abstract: A method and device for model compression of a neural network. The method comprises: recording input and output parameters of each layer of network in a network structure; dividing the network structure into several small networks according to the input and output parameters; setting a pruning flag bit of a first convolutional layer in each small network to be zero to obtain a pruned small network; training each pruned small network to obtain a network weight and a weight mask; recording a pruned channel index number of each convolutional layer of a pruned small network with the weight mask of zero; and carrying out decomposition calculation on each pruned small network according to the pruned channel index number. According to the method, a calculation amount and the size of a model is reduced, and during network deployment, the model can be loaded with one click, thus reducing usage difficulty.

Abstract: Provided is a system for analyzing a multivariate time series that includes at least one processor programmed or configured to receive a time series of historical data points, determine a historical time period, determine a contemporary time period, determine a first time series of data points associated with a historical transaction metric from the historical time period, determine a second time series of data points associated with a historical target transaction metric from the historical time period, determine a third time series of data points associated with a contemporary transaction metric from the contemporary time period, and generate a machine learning model, wherein the machine learning model is configured to provide an output that comprises a predicted time series of data points associated with a contemporary target transaction metric. Methods and computer program products are also provided.

Abstract: A method includes receiving input data including a plurality of feature vectors and labeling each feature vector based on a temporal proximity of the feature vector to occurrence of a fault. Feature vectors that are within a threshold temporal proximity to the occurrence of the fault are labeled with a first label value and other feature vectors are labeled with a second label value. The method includes determining, for each feature vector of a subset, a probability that the label associated with the feature vector is correct. The subset includes feature vectors having labels that indicate the first label value. The method includes reassigning labels of one or more feature vectors of the subset having a probability that fails to satisfy a probability threshold and, after reassigning the labels, training an aircraft fault prediction classifier using supervised training data including the plurality of feature vectors and the labels.

Abstract: One or more computing devices, systems, and/or methods for entity disambiguation are provided. For example, a document may be analyzed to identify a first mention and a second mention. One or more techniques may be used to select and link a candidate entity, from a first set of candidate entities, to the first mention and select and link a candidate entity, from a second set of candidate entities, to the second mention.

Abstract: Some embodiments provide a method for reducing complexity of a machine-trained (MT) network that receives input data and computes output data for each input data. The MT network includes multiple computation nodes that (i) generate output values and (ii) use output values of other computation nodes as input values. During training of the MT network, the method introduces probabilistic noise to the output values of a set of the computation nodes. the method determines a subset of the computation nodes for which the introduction of the probabilistic noise to the output value does not affect the computed output data for the network. The method removes the subset of computation nodes from the trained MT network.

Abstract: In certain embodiments, resolved exceptions information regarding resolved exceptions may be obtained. The resolved exceptions information may indicate the resolved exceptions and, for each resolved exception of the resolved exceptions, a set of attributes of a transaction for which the resolved exception was triggered. The resolved exceptions information may be provided as input to a prediction model to obtain multiple decision trees via the prediction model. Each decision tree of the multiple decision trees may comprise nodes and conditional branches, each node of the nodes of the decision tree indicating a probability of a dividend-related classification for a transaction that corresponds to the node. A decision tree may be obtained from the multiple decision trees.

Abstract: A method may include receiving a communication from a device at an artificial intelligence controller including state information for a software application component running on the device, the state information including information corresponding to at least one potential state change available to the software application component, and metrics associated with at least one end condition, interpreting the state information using the artificial intelligence controller, and selecting an artificial intelligence algorithm from a plurality of artificial intelligence algorithms for use by the software application component based on the interpreted state information; and transmitting, to the device, an artificial intelligence algorithm communication, the artificial intelligence algorithm communication indicating the selected artificial intelligence algorithm for use in the software application component on the device.

Abstract: In an example, an apparatus comprises a plurality of execution units comprising at least a first type of execution unit and a second type of execution unit and logic, at least partially including hardware logic, to expose embedded cast operations in at least one of a load instruction or a store instruction; determine a target precision level for the cast operations; and load the cast operations at the target precision level. Other embodiments are also disclosed and claimed.

Abstract: Provided is an apparatus configured to determine a common neural network based on a comparison between a first neural network included in a first application program and a second neural network included in a second application program, utilize the common neural network when the first application program or the second application program is executed.

Abstract: A graph neural network (GNN)-based prediction system for total organic carbon (TOC) in shale solves the problem that the existing shale TOC prediction method cannot fully analyze the complex nonlinear relationship between all logging curves and TOC. The prediction system adopts a method including: acquiring and preprocessing a plurality of logging curves of a target well location in a target shale bed to acquire a plurality of standardized logging curves, windowing the plurality of standardized logging curves, and inputting the windowed logging curves and weight matrix into a trained GNN-based TOC prediction network to acquire TOC of the target well location. The prediction system inputs the plurality of logging curves as correlative multi-dimensional dynamic graph data for analysis and can acquire the complex nonlinear relationship between the logging curves and TOC, thus improving the prediction accuracy of TOC.

Abstract: Disclosed herein are techniques for implementing a large fully-connected layer in an artificial neural network. The large fully-connected layer is grouped into multiple fully-connected subnetworks. Each fully-connected subnetwork is configured to classify an object into an unknown class or a class in a subset of target classes. If the object is classified as the unknown class by a fully-connected subnetwork, a next fully-connected subnetwork may be used to further classify the object. In some embodiments, the fully-connected layer is grouped based on a ranking of target classes.

Abstract: At a rule processing unit of an evolving, self-organized machine learning-based resource management service, a rule of a first rule set is applied to a value of a first collected metric, resulting in the initiation of a first corrective action. A set of metadata indicating the metric value and the corrective action is transmitted to a repository, and is used as part of an input data set for a machine learning model trained to generate rule modification recommendations. In response to determining that the corrective actions did not meet a success criterion, an escalation message is transmitted to another rule processing unit.

Abstract: Embodiments of the invention include computer-implemented methods, computer systems, and computer program products for predicting sensory perception. A non-limiting example of the computer-implemented method includes receiving at a processor a library including a plurality of indexed sensory descriptors. A sensory target descriptor is also received at the processor. The processor is configured to calculate a coefficient matrix based in part on the semantic distance between an indexed sensory descriptor and a sensory target descriptor. The processor is further configured to generate a perceptual descriptor prediction for the sensory target.

Abstract: Mechanisms for generating different lines of questioning based on evaluation of a previous line of questioning are provided. A set of evidential data specifying a plurality of concept entities and input specifying a scenario to be evaluated are received. The scenario specifies a hypothetical link between at least two of the concept entities. A first set of questions corresponding to the at least two information concept entities are evaluated based on the set of evidential data. Based on results of evaluating the first set of questions, a second set of questions is automatically generated to further expand upon and investigate the results of evaluating the first set of questions. The second set of questions are processed and an indication of the scenario and a corresponding measure of support for or against the scenario being a valid scenario is output.