Abstract: A pseudo-relevance feedback (PRF) system is disclosed that determines an optimized relevance model for a search query by utilizing a posterior relevance model to estimate the likelihood that an initial set of top-K retrieved documents would be retrieved given the posterior relevance model, re-ranking the top-K documents based on their respective estimates of likelihood of retrieval, determining a rank similarity between the initial ranking of the top-K documents and the re-ranking of the top-K documents, updating one or more model parameters of the posterior relevance model based on the rank similarity, and iteratively performing the above process until the rank similarity is maximized, at which point, the optimized relevance model is obtained.

Abstract: A method, system and computer readable medium for performing a cognitive search operation comprising: receiving training data, the training data comprising information based upon user interaction with cognitive attributes; performing a machine learning operation on the training data; generating a cognitive profile based upon the information generated by performing the machine learning operation; and, performing a cognitive search operation on a corpus of content based upon the cognitive profile, the cognitive search operation returning cognitive results specific to the cognitive profile of the user.

Abstract: Input image data having a plurality of pixel values represented in a two-dimensional matrix form of columns and rows is received. The input image data is transformed into a plurality of input rows. The pixel values in each input row correspond to the pixel values in a predetermined subset of the columns of the input image data and all of the rows of each column of the subset of columns. A plurality of subsets of pixel values in the plurality of input rows is determined. The number of pixel values in each row of a subset of pixel values equal in number to a number of filter values in a filter. Each input row of each subset of pixel values is convolved with the filter values of the filter to determine a corresponding output value and stored in a memory.

Abstract: An electronic device is described which has at least one input interface to receive at least one item of a sequence of items. The electronic device is able to communicate with a server, the server storing a neural network and a process which generates item embeddings of the neural network. The electronic device has a memory storing a copy of the neural network and a plurality of item embeddings of the neural network. In the case when there is unavailability at the electronic device of a corresponding item embedding corresponding to the received at least one item, the electronic device triggers transfer of the corresponding item embedding from the server to the electronic device. A processor at the electronic device predicts at least one candidate next item in the sequence by processing the corresponding item embedding with the copy of the neural network and the plurality of item embeddings.

Abstract: A convolution neural network (CNN) model is trained and pruned at a pruning ratio. The model is then trained and pruned one or more times without constraining the model according to any previous pruning step. The pruning ratio may be increased at each iteration until a pruning target is reached. The model may then be trained again with pruned connections masked. The process of pruning, retraining, and adjusting the pruning ratio may also be repeated one or more times with a different pruning target.

Abstract: Methods an systems to classify a training dataset of network data as a poisoned training dataset based on a first dataset-level classifier, identify and remove poison samples of the poisoned training dataset based on a sample-level classifier to produce a non-poisoned dataset, training a machine-based model to analyze network traffic based on the modified non-poisoned dataset, and analyze network traffic with the machine-based model.

Abstract: Systems, apparatuses, and methods are provided for identifying a corresponding string stored in memory based on an incomplete input string. A system can analyze and produce phonetic and distance metrics for a plurality of strings stored in memory by comparing the plurality of strings to an incomplete input string. These similarity metrics can be used as the input to a machine learning model, which can quickly and accurately provide a classification. This classification can be used to identify a string stored in memory that corresponds to the incomplete input string.

Abstract: A computer-implemented method includes employing a dynamic Boltzmann machine (DyBM) to solve a maximum likelihood of generalized normal distribution (GND) of time-series datasets. The method further includes acquiring the time-series datasets transmitted from a source node to a destination node of a neural network including a plurality of nodes, learning, by the processor, a time-series generative model based on the GND with eligibility traces, and, performing, by the processor, online updating of internal parameters of the GND based on a gradient update to predict updated times-series datasets generated from non-Gaussian distributions.

Abstract: An online system receives information describing events corresponding to actions associated with a third party system performed by an individual. The received information describes event types and times at which the events occurred. The online system generates nodes of a directed graph associated with the third party system, in which each node corresponds to an event type. For each event, a node count associated with a node corresponding to the event's type is incremented by the online system. Pairs of consecutively occurring events are identified based on times at which the events occurred and an edge describing each transition from one event to another is generated by the online system. The online system determines an edge count for each transition indicating a number of edges describing the transition as well as a sequential order of event types based on one or more node counts and one or more edge counts.

Abstract: A neuromorphic synapse array is provided which ensures that a neuron model as such McCulloch-Pitts is dependent on nonlinearity with a single polarity weight cell. The neuromorphic synapse array includes a plurality of synaptic array cells, a plurality of operation column arrays, and a reference column array. The synaptic array cells respectively have a single polarity synapse weight and are classified into operation synapse cells and reference synapse cells for shifting a product-sum of the operation synapse cells. The operation column arrays are defined by the operation synapse cells aligned in column of the array. The reference column array is defined by the reference synapse cells aligned in column of the array.

Abstract: Embodiments of the invention provide a method comprising maintaining a library of one or more compositional prototypes. Each compositional prototype is associated with a neurosynaptic program. The method further comprises searching the library based on one or more search parameters. At least one compositional prototype satisfying the search parameters is selected. A neurosynaptic network is generated or extended by applying one or more rules associated with the selected compositional prototypes.

Abstract: Systems, methods, and other embodiments are disclosed for partitioning models in a database. In one embodiment, a set of training data is parsed into multiple data partitions based on partition keys, where the data partitions are identified by the partition keys and are used for training data mining models. The multiple data partitions are analyzed to generate partition metrics data. Algorithm data, identifying at least one algorithm for processing the multiple data partitions, and resources data, identifying available modeling resources for processing the multiple data partitions, are read. The partition metrics data, the algorithm data, and the resources data are processed to generate an organization data structure. The organization data structure is configured to control distribution and processing of the multiple data partitions across the available modeling resources to generate a composite model object that includes a separately trained data mining model for each partition of the multiple partitions.

Abstract: A processing unit can operate a first recurrent computational model (RCM) to provide first state information and a predicted result value. The processing unit can operating a first network computational model (NCM) to provide respective expectation values of a plurality of actions based at least in part on the first state information. The processing unit can provide an indication of at least one of the plurality of actions, and receive a reference result value, e.g., via a communications interface. The processing unit can train the first RCM based at least in part on the predicted result value and the reference result value to provide a second RCM, and can train the first NCM based at least in part on the first state information and the at least one of the plurality of actions to provide a second NCM.

Abstract: A system can obtain observations from a dataset. The system can generate a set of training partitions based on the observations and generate an ensemble of machine-learning models based on the set of training partitions. The system can then receive new data and detect whether the new data is indicative of the event using the ensemble. In some cases, the system can update the ensemble by providing the new data as input to an unsupervised machine-learning model that is separate from the ensemble of machine-learning models; receiving an output from the unsupervised machine-learning model indicating whether or not the new data is indicative of the event; incorporating a new observation into the dataset indicating whether or not the new data is indicative of the event based on the output from the unsupervised machine-learning model; and updating the ensemble based on the dataset with the new observation.

Abstract: Quantum neural nets, which utilize quantum effects to model complex data sets, represent a major focus of quantum machine learning and quantum computing in general. In this application, example methods of training a quantum Boltzmann machine are described. Also, examples for using quantum Boltzmann machines to enable a form of quantum state tomography that provides both a description and a generative model for the input quantum state are described. Classical Boltzmann machines are incapable of this. Finally, small non-stoquastic quantum Boltzmann machines are compared to traditional Boltzmann machines for generative tasks, and evidence presented that quantum models outperform their classical counterparts for classical data sets.

Abstract: A method for receiving training data for training a neural network (NN) to perform a machine learning (ML) task and for determining, using the training data, an optimized NN architecture for performing the ML task is described. Determining the optimized NN architecture includes: maintaining population data comprising, for each candidate architecture in a population of candidate architectures, (i) data defining the candidate architecture, and (ii) data specifying how recently a neural network having the candidate architecture has been trained while determining the optimized neural network architecture; and repeatedly performing multiple operations using each of a plurality of worker computing units to generate a new candidate architecture based on a selected candidate architecture having the best measure of fitness, adding the new candidate architecture to the population, and removing from the population the candidate architecture that was trained least recently.

Abstract: In an embodiment, a method for optimizing computer machine learning includes receiving an optimization goal. The optimization goal is used to search a database of base option candidates (BOC) to identify matching BOCs that at least in part matches the goal. A selection of a selected base option among the matching BOCs is received. Machine learning prediction model(s) are selected based at least in part on the goal to determine prediction values associated with alternative features for the selected base option, where the model(s) were trained using training data to at least identify weight values associated with the alternative features for models. Based on the prediction values, at least a portion of the alternative features is sorted to generate an ordered list. The ordered list is provided for use in manufacturing an alternative version of the selected base option with the alternative feature(s) in the ordered list.

Abstract: Disclosed herein are methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating an action selection policy of an execution device for completing a task in an environment. The method includes computing a hybrid sampling policy at a state of the execution device based on a sampling policy and an exploration policy, wherein the exploration policy specifies a respective exploration probability corresponding to each of multiple possible actions in the state, wherein the exploration probability is negatively correlated with a number of times that the each of the multiple possible actions in the state has been sampled; sampling an action among the multiple possible actions in the state according to a sampling probability of the action specified in the hybrid sampling policy; and updating an action selection policy in the state by performing Monte Carlo counterfactual regret minimization based on the action.

Abstract: A method of performing time series prediction by improper learning comprising calculating a plurality of filters based on a symmetric matrix and generating a mapping term based on a time series input and a function. The method may include comprising iteratively: transforming the function using the calculated plurality of filters; predicting an interim output using the transformed function and the mapping term; computing an error of the interim output based on a known output; and updating the mapping term based on the computed error. The method may include generating the mapping term through iterations over a predetermined interval and performing a time series prediction using the mapping term generated over the iterations.

Abstract: The present disclosure describes a system, method, and computer program for automatically classifying user accounts within an entity's computer network, using machine-based-learning modeling and keys from an identity management system. A system uses supervised machine learning to create a statistical model that maps individual keys or sets of keys to a probability of being associated with a first type of user account (e.g., a service account). To classify an unclassified user account, the system identifies identity management keys associated with the unclassified user account. The system creates an N-dimensional vector from the keys (where N=the number of keys), and uses the vector and the statistical model to calculate a probability that the unclassified user account is the first type of user account. In response to the probability exceeding a first threshold, the system classifies the unclassified user account as the first type of user account.