Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium. In one aspect, a method includes the actions of receiving a request to perform computations for a neural network on a hardware circuit having a matrix computation unit, the request specifying a transpose operation to be performed on a first neural network matrix; and generating instructions that when executed by the hardware circuit cause the hardware circuit to transpose the first neural network matrix by performing first operations, wherein the first operations include repeatedly performing the following second operations: for a current subdivision of the first neural network matrix that divides the first neural network matrix into one or more current submatrices, updating the first neural network matrix by swapping an upper right quadrant and a lower left quadrant of each current submatrix, and subdividing each current submatrix into respective new submatrices to update the current subdivision.

Abstract: Computational methods and systems that collect operational data from an intelligent controller to identify information, or correct information, about a device and system controlled by the intelligent controller are disclosed. Computational methods and systems use a set of operational data and information known about other devices and systems controlled by similar intelligent controllers to process the operational data and generate information, or correct information, about the device and system.

Abstract: A method of website optimization including collecting data for constructing user profiles; constructing the user profiles; quantifying affinities between the user profiles; constructing a user network in which the affinities are represented as links between user nodes; constructing an adjacency matrix; calculating a first principal eigenvector of the adjacency matrix; defining a new network by removing a random link in the network and calculating a new adjacency matrix for the new network; calculating a second principal eigenvector of the new adjacency matrix; calculating a vector of relative shifts between the first principal eigenvector and the second principal eigenvector; for every node, assigning a value for a direction of its shift between the network and the new network; and repeating certain of the steps.

Abstract: A computer generates labels for machine learning algorithms by retrieving, from a data storage circuit, multiple label sets that contain labels that each classify data points in a corpus of data. A graph is generated that includes a plurality of edges, each edge between two respective labels from different label sets of the multiple label sets. Weights are determined for the plurality of edges based upon a consistency between data points classified by two labels connected by the edges. An algorithm is applied that groups labels from the multiple label sets based upon the weights for the plurality of edges. Data points are identified from the corpus of data that represent conflicts within the grouped labels. An electronic message is transmitted in order to present the identified data points to entities for further classification. A new label set is generated using the further classification received from the entities.

Abstract: Systems, methods, and other embodiments are disclosed that are configured to automatically complete the generation of task structures in a project plan. In one embodiment, user data is generated by monitoring interactions of a user with a computer user interface when forming a task structure in a project plan. The user data is compared to learned behavior data. The learned behavior data represents reusable patterns derived from information associated with user creation of previous task structures. If, based on the comparison, the user data matches a reusable pattern of the learned behavior data, a prompt is displayed to the user indicating an option to accept or reject automated completion of the task structure. In response to receiving an acceptance from the user, the forming of the task structure is automatically completed in the project plan based on the matching reusable pattern.

Abstract: An enterprise management system is described for efficient operating system migration, preserving applications, data, and settings. A staging area, such as an empty folder, is created on a client device. A base layer for the new operating system and application layers for applications that will be installed on the computing device are downloaded to the staging area. After the base layer and application layers are downloaded, the layers are merged onto the computing device to instantly install the operating system and the applications. User settings, data, and other applications can be migrated to corresponding locations in the new operating system from the old operating system.

Abstract: A method for predicting consumer power demand uses power consumption data measured over a long term and a power usage pattern immediately before a target time and for controlling ESS charge/discharge of an ESS based on the predicted power demand. A power demand prediction apparatus using the method includes respective components for collecting weather data and data on power used by the consumer; selecting data points according to a preset condition from among data collected by the data collector based on a specific time span; generating long term prediction data for the power demand in the specific time span based on the selected data points; analyzing a power usage pattern immediately before the specific time span and comparing the power usage pattern with the long-term prediction data, to determine whether prediction data correction is required; and correcting the prediction data based on the power usage pattern when correction is required.

Abstract: Systems and methods for determining uniqueness of device identifiers are provided are provided. The uniqueness of a device identifier may be indicated by a device quality score or grade that is calculated based on a plurality of parameters associated with a device identifier as well as evaluation rules derived based on historical data. The plurality of parameters may be associated with a network event or transaction associated with the device identifier. The evaluation rules may be derived using machine learning techniques. Based on uniqueness of a device identifier, a suitable action or measure may be taken.

Abstract: Systems for distributed data storage. A method embodiment commences upon capturing a history of storage I/O activity over a recent time period. A predictive model is derived from the captured storage I/O activity, and the predictive model is then used for predicting future storage I/O activity. A set of snapshot planning parameters comprising objectives (e.g., to minimize costs or to maximize likelihood completing a snapshot activity by a prescribed time) and/or constraints (e.g., don't wait more than one day to start a snapshot) are applied to the predicted storage I/O characteristics to generate a set of feasible snapshot plans. One of the feasible snapshot plans is selected for scheduling so as to begin the planned snapshot activity at a prescribed time. The snapshot planning parameters are normalized based on the predicted storage I/O characteristics.

Abstract: A first list that includes a first set of one or more concept terms may be obtained. The first set of one or more concept terms may be candidates for being conceptually related to a seed concept term. Prior to the obtaining, a first client computing device may provide a first plurality of individual characters and the seed concept term to a first user using the first client computing device. The first client computing device may prompt the first user to generate the first set of one or more concept terms that are conceptually related to the seed concept term using one or more of the first plurality of individual characters. In response to the obtaining, one or more particular categories to which each of the first set of one or more concept terms belongs to may be determined.

Abstract: Techniques that facilitate machine learning using multi-dimensional time series data are provided. In one example, a system includes a snapshot component and a machine learning component. The snapshot component generates a first sequence of multi-dimensional time series data and a second sequence of multi-dimensional time series data from multi-dimensional time series data associated with at least two different data types generated by a data system over a consecutive period of time. The machine learning component that analyzes the first sequence of multi-dimensional time series data and the second sequence of multi-dimensional time series data using a convolutional neural network system to predict an event associated with the multi-dimensional time series data.

Abstract: Systems and methods for training a neural network or an ensemble of neural networks are described. A hyper-parameter that controls the variance of the ensemble predictors is used to address overfitting. For larger values of the hyper-parameter, the predictions from the ensemble have more variance, so there is less overfitting. This technique can be applied to ensemble learning with various cost functions, structures and parameter sharing. A cost function is provided and a set of techniques for learning are described.

Abstract: A method, a computer-readable medium, and an apparatus for a sparse neural network are provided. The apparatus may include a hardware accelerator. The apparatus may determine, for each pair of operands to be processed by a MAR unit, whether both operands of the pair are non-zero. The apparatus may prevent a pair of operands to be processed by the MAR unit from being loaded to a multiplier of the MAR unit when an operand of the pair of operands is zero. The apparatus may place the pair of operands into one of a plurality of queues when both operands of the pair of operands are non-zero.

Abstract: Embodiments of the invention are directed to methods and devices for predicting interactions. One embodiment is directed to a method comprising receiving, by one or more computers, interaction data for a plurality of known interactions between resource providers and users, and creating a topological graph based on the plurality of known interactions. The method may further comprise determining, by the one or more computers, a plurality of communities to form a predictive model, and receiving a request for a prediction. In addition, the method may comprise applying the request to the predictive model, by the one or more computers, by identifying a community in the plurality of communities corresponding to the request, determining a node within the identified community, and providing information regarding the node as the requested prediction.

Abstract: In order to compare high-dimensional, multi-modal data for a patient to data for other patients, deep learning is used to encode original, multi-modal data for a patient into a compact signature. The compact signature is compared to predetermined compact signatures generated for other patients, and similar predetermined compact signatures are identified based on the comparison. A clinical outcome may be predicted based on the similar predetermined compact signatures that are identified.

Abstract: A dynamic power budget allocation system includes a plurality of powered subsystems. A power system controller is coupled to the plurality of powered subsystems. The power system controller is operable, for each of a plurality of time intervals, to retrieve power usage data from each of the plurality of subsystems during a current time interval. The power system controller is then operable to project power requirements for the plurality of subsystems for a subsequent time interval using the power usage data. The power system controller is then operable to determine at least one power setting for at least one of the plurality of subsystems using the power requirements, and program the at least one of the plurality of subsystems with the at least one power setting. Each powered subsystem may include a voltage regulator that provides the power usage data and is programmed with the at least one power setting.

Abstract: Executing a graph. A method includes identifying that a graph includes one or more nodes having input ports that are coupled to alternative inputs. The method further includes identifying permutations of inputs where each permutation uses one of the alternative inputs. The method further includes executing the graph for each of the identified permutations. The method further includes identifying a metric for each execution of the graph for each of the permutations. The method further includes providing a user with an indication of the metrics.

Abstract: Methods and apparatus to facilitate dynamic classification for market research are disclosed. Example disclosed methods include constructing, using a programmed processor based on data for a sample population and a first set of input variables, a self-organizing map classifying the sample population according to a plurality of classes defined in the map using fuzzy class membership. Example disclosed methods include extracting the fuzzy class membership for the sample population from the map. Example disclosed methods include correlating fuzzy class membership with behavior data for the sample population to determine a likely class behavior for the plurality of classes. Example disclosed methods include using fuzzy class membership and the likely class behavior to provide a predictive market output in response to a query.

Abstract: A neuromorphic device for the analog computation of a linear combination of input signals, for use, for example, in an artificial neuron. The neuromorphic device provides non-volatile programming of the weights, and fast evaluation and programming, and is suitable for fabrication at high density as part of a plurality of neuromorphic devices. The neuromorphic device is implemented as a vertical stack of flash-like cells with a common control gate contact and individually contacted source-drain (SD) regions. The vertical stacking of the cells enables efficient use of layout resources.

Abstract: A method and apparatus of a device that incorporates a user preference into a multi-objective optimization while the multi-objective optimization is running is described. In an exemplary embodiment, the device generates a first plurality of intermediate designs based on optimizing a plurality of variable values corresponding to the multiple dimensions of the multi-objective optimization. In addition, each of the first plurality of intermediate designs includes a corresponding weight. Furthermore, the device outputs the first plurality of intermediate designs. The device additionally receives a preference indicator for a selected one of the intermediate designs, where a user inputs the preference indicator while the multi-objective optimization is running. The device further adjusts the corresponding weight of the selected one of the first plurality of intermediate designs based on the preference indicator.