Abstract: Systems, apparatuses and methods may provide for a chip that includes a memory array having a plurality of rows corresponding to neurons in a spiking neural network (SNN) and a row decoder coupled to the memory array, wherein the row decoder activates a row in the memory array in response to a pre-synaptic spike in a neuron associated with the row. Additionally, the chip may include a sense amplifier coupled to the memory array, wherein the sense amplifier determines post-synaptic information corresponding to the activated row. In one example, the chip includes a processor to determine a state of a plurality of neurons in the SNN based at least in part on the post-synaptic information and conduct a memory array update, via the sense amplifier, of one or more synaptic weights in the memory array based on the state of the plurality of neurons.

Abstract: A non-transitory computer-readable medium including instructions, which when executed by one or more processors of a computing system, causes the computing system to: access a machine learning model m, an input data point P to m, P including one or more features, and a prediction m(P) of m for P; create a set of perturbed input data points Pk from P by selecting a new value for at least one feature of P for each perturbed input data point; obtain a prediction m(Pk) for each of the perturbed input data points; analyze the predictions m(Pk) for the perturbed input data points to determine which features are most influential to the prediction; and output the analysis results to a user.

Abstract: Computer-implemented systems and methods soft-tie learned parameters of a neural network(s). The soft-tying comprises: applying a common label to the first and second learned parameters; and as part of the training, and in response to the first and second learned parameters having the common label, applying a regularization penalty to a loss function for the first learned parameter upon a determination that the first learned parameter is different than the second learned parameter. The learned parameters can be connection weights, node biases, and/or parametric model statistics. The application of the regularization penalty can be influenced by a soft-tying hyperparameter.

Abstract: Systems and techniques for neural network training are described herein, a training set may be received for a neural network. Here, the neural network may comprise a set of nodes arranged in layers and a set of inter-node weights between nodes in the set of nodes. The neural network may then be iteratively trained to create a trained neural network. An iteration of the training may include generating a random unit vector and creating an update vector by calculating a magnitude for the random unit vector based on a degree that the random unit vector matches a gradient—where the gradient is represented by a dual number. The iteration may continue by updating a parameter vector for an inter-node weight by subtracting the update vector from a previous parameter vector of the inter-node weight. The trained neural network may then be used to classify data.

Abstract: An artificial neural network device that utilizes one or more non-volatile memory arrays as the synapses. The synapses are configured to receive inputs and to generate therefrom outputs. Neurons are configured to receive the outputs. The synapses include a plurality of memory cells, wherein each of the memory cells includes spaced apart source and drain regions formed in a semiconductor substrate with a channel region extending there between, a floating gate disposed over and insulated from a first portion of the channel region and a non-floating gate disposed over and insulated from a second portion of the channel region. Each of the plurality of memory cells is configured to store a weight value corresponding to a number of electrons on the floating gate. The plurality of memory cells are configured to multiply the inputs by the stored weight values to generate the outputs.

Abstract: Systems, methods, and storage media for training a machine learning model are disclosed. Exemplary implementations may select a set of training images for a machine learning model, extract object features from each training image to generate an object tensor for each training image, extract stylistic features from each training image to generate a stylistic feature tensor for each training image, determine an engagement metric for each training image, and train a neural network comprising a plurality of nodes arranged in a plurality of sequential layers.

Abstract: Techniques that facilitate time series analysis using machine learning are provided. In one example, a system includes a matrix generation component, a matrix factorization component and a machine learning component. The matrix generation component converts at least a first stream of time series data and a second stream of time series data (e.g., raw time series data) into a data matrix (e.g., a partially-observed similarity matrix) that comprises void data and numerical data associated with the first stream of time series data and the second stream of time series data. The matrix factorization component factorizes the data matrix into a first factorization data matrix and a second factorization data matrix. The machine learning component processes a machine learning model based on first matrix data associated with the first factorization data matrix and second matrix data associated with the second factorization data matrix.

Abstract: Semantic information that describes data sets is inferred based upon a semantic analysis performed on data sets retained within a data repository. The semantic analysis can include a determination of formats associated with fields of the data sets and a comparison of values of the fields against reference data sets having predetermined semantic types. Correlations are inferred between data sets based upon respective semantic information. The correlations are incorporated into visualizations displayed in connection with a graphical user interface.

Abstract: A technique for answering questions includes receiving a question directed to a first subject. A mathematical operation is performed between each of one or more first topic vectors (associated with the first subject) and each of one or more second topic vectors (associated with a second subject) to generate respective strength values. Relevant ones of the respective strength values are summed to provide an overall strength value, which is utilized to determine a semantic distance (SD) between the first subject and the second subject. In response to the SD being within a threshold distance value (TDV), information associated with the first subject and the second subject is utilized to answer the question. In response to the SD not being within the TDV, information associated with the first subject is utilized to answer the question.

Abstract: A system, a machine-readable storage medium storing instructions, and a computer-implemented method as described herein are directed to a Personalized Article Engine that generates respective prediction models for each article in a plurality of candidate articles in a social network system. The Personalized. Article Engine generates a respective article score according to each article's prediction model and at least one feature of a target member account. The Personalized Article Engine generates a plurality of output scores based on combining each respective article score with a corresponding article's global model score. The Personalized Article Engine ranks the output scores to identify a subset of candidate articles relevant to the target member account.

Abstract: A machine learning model can be trained to infer the probability of the presence of categories of a software bug in a source code file. A bug tracker can provide information concerning the category to which a software bug belongs. The bug data supplied to a machine learning model for inferring the presence of particular categories of bugs can be filtered to exclude a specified category or categories of bugs. Information including but not limited to organizational boundaries can be inferred from the category of bugs present in a body of source code. The inferred organization boundaries can be used to generate team-specific machine learning models.

Abstract: Techniques that facilitate time series analysis using machine learning are provided. In one example, a system includes a matrix generation component, a matrix factorization component and a machine learning component. The matrix generation component converts at least a first stream of time series data and a second stream of time series data (e.g., raw time series data) into a data matrix (e.g., a partially-observed similarity matrix) that comprises void data and numerical data associated with the first stream of time series data and the second stream of time series data. The matrix factorization component factorizes the data matrix into a first factorization data matrix and a second factorization data matrix. The machine learning component processes a machine learning model based on first matrix data associated with the first factorization data matrix and second matrix data associated with the second factorization data matrix.

Abstract: Device configuration based on predicting a health affliction. A process acquires measurements of conditions that a user is experiencing. The process predicts, based on the measurements, whether the user will experience a particular health affliction. Based on predicting that the user will experience the particular health affliction, the process configures devices of an environment in which the user is present to reduce effects of the devices on symptoms of the particular health affliction. The configuring includes adjusting a respective at least one state of each device of the devices.

Abstract: VQE is accelerated by performing receiving a qubit Hamiltonian representing a linear combination of a plurality of Pauli strings. Selecting, among the plurality of Pauli strings, one or more Pauli strings that have less influence than a threshold on an eigenvalue of the qubit Hamiltonian. Grouping, based on joint measurability, the unselected Pauli strings among the plurality of Pauli strings into a plurality of groups of jointly measurable Pauli strings Determining that one or more of the selected one or more Pauli strings is jointly measurable with Pauli strings in one of the plurality of groups And adding one or more of the selected one or more Pauli strings to the one of the plurality of groups.

Abstract: Technologies for automatic discovery and connection to a representational state transfer (REST) interface include a provider computing device communicatively coupled to a REST interface of a Web service hosted by a 3rd party. The provider computing device is configured to analyze a data representation received from a REST interface of a Web service in response to having transmitted an HTTP request to an endpoint of the Web service and determine a pattern of the data representation as a function of the analysis of the data representation. Additionally, the provider computing device is configured to generate one or more possible schemas for the REST interface based on the determined pattern. Additional embodiments are described herein.

Abstract: In one aspect, this specification describes a recurrent neural network system implemented by one or more computers that is configured to process input sets to generate neural network outputs for each input set. The input set can be a collection of multiple inputs for which the recurrent neural network should generate the same neural network output regardless of the order in which the inputs are arranged in the collection. The recurrent neural network system can include a read neural network, a process neural network, and a write neural network. In another aspect, this specification describes a system implemented as computer programs on one or more computers in one or more locations that is configured to train a recurrent neural network that receives a neural network input and sequentially emits outputs to generate an output sequence for the neural network input.

Abstract: The present disclosure provides an artificial neural network communicatively-coupled to at least one computer having one or more processors, including a plurality of neurons arranged in layers. The artificial neural network is arranged to receive a new neuron into a layer of the artificial neural network during training; the new neuron is added to the neural network when no other neuron in that layer for a selected output can learn a relationship associated with an input vector of a data set being learnt. The new neuron is updated with both the relationship which could not be learnt by any other neuron in that layer and a modified data set from a last trained neuron in that layer that contributes to the selected output of the neural network. Methods and computer-readable media are also disclosed.

Abstract: A scalable system and method for completing a model task using a serverless architecture is disclosed. The system may include a model optimizer having one or more memory units for storing instructions and one or more processors. The method may include receiving a request to complete a model task, and retrieving a stored model and a first hyperparameter based on the request. The method may include provisioning first computing resources to a development instance configured to train the retrieved model based on the first hyperparameter and the model task. The method may include receiving, from the development instance, a trained model and a performance metric. The method may include receiving, from a different development instance, a different performance metric associated with a different model, and terminating the development instance based on a determination that the termination condition is satisfied.

Abstract: Methods, systems, and apparatus for solving optimization tasks. In one aspect, a system includes one or more classical processors and one or more quantum computing resources, wherein the one or more classical processors and one or more quantum computing resources are configured to perform operations comprising receiving input data comprising data specifying a computational task to be solved; processing the received input data using a first quantum computing resource to generate data representing a reduced computational task, wherein the reduced computational task has lower dimensionality that the computational task; and processing the data representing the reduced computational task to obtain a solution to the computational task.

Abstract: A method, system and computer-usable medium are disclosed for automating the generation of an incorrect answer to a question suitable for a multiple choice exam. An input corpus of human-readable text associated with a subject domain is provided to a question generation system, where it is processed to generate a set of question-answer (QA) pairs. The set of QA pairs is then processed with the corpus of input text to extract a set of input keywords and concepts. A concept dependency graph is then used to perform disambiguation operations on the set of input keywords and concepts, and the reference keywords and concepts it contains, to generate a set of distractor words. The resulting set of distractor words is then processed with the set of QA pairs to generate a set of multiple choice question-answers that include various distractor answers.