Abstract: Systems and methods for reducing the size of deep neural networks are disclosed. In an embodiment, a server computer stores a plurality of training datasets, each of which comprise a plurality of training input matrices and a plurality of corresponding outputs. The server computer initiates training of a deep neural network using the plurality of training input matrices, a weight matrix, and the plurality of corresponding outputs. While the training of the deep neural network is being performed, the server computer identifies one or more weight values of the weight matrix for removal. The server computer removes the one or more weight values from the weight matrix to generate a reduced weight matrix. The server computer then stores the reduced weight matrix with the deep neural network.

Abstract: Methods, systems, and apparatus for solving optimization tasks. In one aspect, a method includes receiving input data comprising (i) data specifying an optimization task to be solved, and (ii) data specifying task objectives for solving the optimization task, comprising one or more local task objectives and one or more global task objectives; processing the received input data to obtain one or more initial solutions to the optimization task based on the local task objectives, wherein at least one initial solution is obtained from a first quantum computing resource; and processing the generated one or more initial solutions using a second quantum computing resource to generate a global solution to the optimization task based on the global task objectives.

Abstract: A wire electric discharge machine according to the present invention includes a machine learning device which performs machine learning for adjustment of a machining condition of the wire electric discharge machine, the machine learning device includes a state observation unit which acquires data related to a machining state of a workpiece, a reward calculation unit which calculates a reward based on data related to a machining state, a machining condition adjustment learning unit which determines an adjustment amount of a machining condition based on a machine learning result and data related to a machining state, and a machining condition adjustment unit which adjusts a machining condition based on the determined adjustment amount of a machining condition, and the machining condition adjustment learning unit performs machine learning for adjustment of a machining condition based on the determined adjustment amount of a machining condition, data related to a machining state and acquired by the state observat

Abstract: A plurality of synapse determination circuits are provided on a one-to-one basis for a plurality of gate electrodes of a multi-input gate electrode in a neuron element. With respect to first image regions where “1” is repeatedly inputted in correspondence with group information, the synapse determination circuits corresponding to the first image regions are excitatory synapses. With respect to second image regions where “0” is repeatedly inputted in correspondence with the group information, the synapse determination circuits corresponding to the second image regions are inhibitory synapses.

Abstract: A machine learning apparatus includes a state observing unit for observing a state variable comprised of at least one of an actual dimension value, a resistance actual value, etc., and at least one of a dimension command value, a resistance command value, etc., and an execution time command value for a program, and a learning unit for performing a learning operation by linking at least one of an actual dimension value, a resistance actual value, etc., to at least one of a dimension command value, a resistance command value, etc., observed by the state observing unit, and an execution time command value for the program.

Abstract: A control device in a vehicle includes a unit for calculating, during operation of the vehicle, on the basis of at least one input variable ascertained during operation, at least one output variable for a control system of functions of the vehicle. The control device performs the calculation of the output variables using a Bayesian regression of training values ascertained, before operation, for the output variable and the input variable.

Abstract: A machine learning apparatus includes a state observing unit and a learning unit. The state observing unit observes a state variable comprised of at least one of an adhesion state, a dielectric strength voltage, an electric heating time temperature, and an actual electric heating time value of a coil electrically heated by a coil electric heating unit, and at least one of an electric heating time command value, a voltage, and a current in the coil electric heating unit. The learning unit performs a learning operation by linking at least one of an adhesion state, a dielectric strength voltage, an electric heating time temperature, and an actual electric heating time value of the coil observed by the state observing unit to at least one of the electric heating time command value, the voltage, and the current, which are observed by the state observing unit.

Abstract: A machine learning apparatus that learns a condition associated with a gain of a magnetic flux controller and a time constant of a magnetic flux estimator in a motor control apparatus includes: a state observation unit that observes a state variable defined by at least one of data relating to an acceleration of a motor, data relating to a jerk of the motor, and data relating to an acceleration time of the motor; and a learning unit that learns the condition associated with the gain of the magnetic flux controller and the time constant of the magnetic flux estimator in accordance with a training data set defined by the state variable.

Abstract: A method and apparatus for providing personalized configuration of physical supports for the human body, comprising accepting input including an individual's demographic information, neurological attributes, physical history, operational environment, and outcome or use objectives, processing user input employing an artificial intelligence engine, and then returning guidance and/or control parameters directed to seating adjustment and positioning, including incline angles for wheelchair tilt and recline.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.

Abstract: Apparatus, systems, and methods for analyzing data are described. The data can be analyzed using a hierarchical structure. One such hierarchical structure can comprise a plurality of layers, where each layer performs an analysis on input data and provides an output based on the analysis. The output from lower layers in the hierarchical structure can be provided as inputs to higher layers. In this manner, lower layers can perform a lower level of analysis (e.g., more basic/fundamental analysis), while a higher layer can perform a higher level of analysis (e.g., more complex analysis) using the outputs from one or more lower layers. In an example, the hierarchical structure performs pattern recognition.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.

Abstract: The modeling of an impression effect may include generating a content item impression effect distribution. A classification model may be used to determine a period of the content item impression effect distribution based on one or more accessed impression effect parameters. A value for a content item may be determined based, at least in part, on the determined period and a bid associated with the content item. A content item may be selected based on the determined value and data to display the selected content item may be transmitted. In some instances, the determined period may be used to determine or select predictive model for the determined period that outputs a factor to modify the determined value.

Abstract: Disclosed are various embodiments for using contrarian machine learning models to compensate for selection bias. Both a primary machine learning model and a contrarian machine learning model may be trained for selecting sets of items based at least in part on the same training data. However, the contrarian machine learning model is specially trained to avoid selecting items that are selected by the primary machine learning model. Items selected by the primary model and items selected by the contrarian model are presented to users as recommendations. Both models are updated based at least in part on user selections of items. Ultimately, the use of the contrarian model avoids causing the primary model to degenerate to picking random items due to reinforcement resulting from a bias in favor of selecting items that have been recommended.

Abstract: According to one exemplary embodiment, a method for solving combinatorial optimization problems is provided. The method may include receiving a plurality of problem instance parameters associated with a graph. The method may also include determining a dynamic path change indicator exists. The method may then include initializing the graph based on the determining the dynamic path change indicator does not exist. The method may further include inserting a placeholder node and at least one placeholder node edge based on the determining the dynamic path change indicator exists. The method may also include reinitializing the graph with the inserted place holder node and the at least one placeholder node edge. The method may then include initializing the reinitialized graph. The method may further include executing a hybrid algorithm on the initialized graph or on the reinitialized graph, wherein the hybrid algorithm comprises an ant colony optimization algorithm and a genetic algorithm.

Abstract: According to one exemplary embodiment, a method for solving combinatorial optimization problems is provided. The method may include receiving a plurality of problem instance parameters associated with a graph. The method may also include determining a dynamic path change indicator exists. The method may then include initializing the graph based on the determining the dynamic path change indicator does not exist. The method may further include inserting a placeholder node and at least one placeholder node edge based on the determining the dynamic path change indicator exists. The method may also include reinitializing the graph with the inserted place holder node and the at least one placeholder node edge. The method may then include initializing the reinitialized graph. The method may further include executing a hybrid algorithm on the initialized graph or on the reinitialized graph, wherein the hybrid algorithm comprises an ant colony optimization algorithm and a genetic algorithm.

Abstract: Provided is a method and apparatus for managing failure modes for condition based maintenance in marine resource production equipment.

Abstract: In one embodiment, network data is received at a Learning Machine (LM) in a network. It is determined whether the LM recognizes the received network data based on information available to the LM. When the LM fails to recognize the received network data: a connection to a central management node is established, a request is sent for information relating to the unrecognized network data to the central management node, and information is received from the central management node in response to the request. The received information assists the LM in recognizing the unrecognized network data.

Abstract: A cognitive system that automatically assesses subjective answers may be provided. A cognitive engine executing on one or more processors may determine whether a statement parsed from a subjective answer by natural language processing technique is accurate or inaccurate, for each of the plurality of statements based on matching the statement with information associated with a domain of a question from a plurality of data sources, according to an accuracy threshold. An overall assessment of the answer may be automatically determined based on a number of statements determined to be accurate, a number of statements determined to be inaccurate, a number of duplicate statements in the answer relative to a total number of statements in the answer. A visual graphics representing accurate and inaccurate statements may be presented or displayed, allowing a user to interact with the visual graphics to modify the assessment.

Abstract: An inference task is performed using a computation device having a plurality of processing elements operable in parallel and connected via a connectivity system. Performing the task includes accepting at the device a specification of at least part of the inference task. The specification characterizes a plurality of variables and a plurality of factors, each factor being associated with a subset of the variables. Each of the processing elements is configured with data defining one or more of the plurality of factors. At each of the processing elements, computation associated with one of the factors is performed concurrently with other of the processing elements performing computation associated with different ones of the factors. Messages are exchanged via a connectivity system. The messages provide inputs and/or outputs to the processing elements for the computations associated with the factors and provide a result of performing of the at least the part of the inference task.