Abstract: Techniques are described herein for provided for augmenting graph networks. Upon receiving a request from an edge device, a graph system prompts a generative model and compares its output to a first graph network of nodes and edges based on similarity scores and degrees of separation. If an adequate response is not found, the graph system identifies and evaluates a second set of nodes outside the first graph network, extracts relevant metadata, and creates a second graph network with new relationships. Contextual natural language is generated from the metadata to form a response, which is returned to the edge device, and may involve controlling one or more assets or devices in response to the context of the request and/or the response.

Abstract: Neuron circuits are provided for spiking neural network apparatus having multiple such neuron circuits interconnected by links, each associated with a respective weight, for transmission of signals between neuron circuits. A neuron circuit includes a digital transmitter for generating trigger signals, indicating a state of the neuron circuit, on outgoing links of the circuit. The state is encoded in a time interval defined by these trigger signals. The neuron circuit includes a digital receiver for detecting such trigger signals on incoming links of the circuit, and digital accumulator logic. In response to detecting a trigger signal on an incoming link, the digital accumulator logic is adapted to generate a weighted signal dependent on the time interval and to accumulate the weighted signals generated from trigger signals on the incoming links to determine the state of the neuron circuit.

Abstract: Systems, methods, and computer program products train a residual neural network including a first fully connected layer, a first recurrent neural network layer, and at least one skip connection for anomaly detection. The at least one skip connection directly connects at least one of (i) an output of the first fully connected layer to a first other layer downstream of the first recurrent neural network layer in the residual neural network and (ii) an output of the first recurrent neural network layer to a second other layer downstream of a second recurrent neural network layer in the residual neural network.

Abstract: Communication-capable devices such as commercial Wi-Fi devices can be used for integrated sensing and communications (ISAC) systems to jointly exchange data and monitor environment. Such devices typically require diverse signal processing such as machine learning inference that demands high-power operations for real-time sensing and computing. The present invention provides a way to realize energy-efficient computing by exploiting the capability of data communications to access distributed computing resources including classical computers and quantum computers over networks. The system and method are based on the realization that computationally intensive processing is offloaded to networked hybrid classical-quantum computing to build dynamic computing graphs. Some embodiments use automated classical-quantum machine learning whose circuits and hyperparameters are automatically adjusted via gradient or heuristic optimization for Wi-Fi indoor monitoring and human tracking.

Abstract: A computer-implemented approach for integrating quantum computing elements into a neural network architecture, the neural network including an encoder and a decoder, the encoder being used to encode data input into the neural network and the decoder being used to at least partially reconstruct the encoded data. The encoder features at least one layer made up of quantum-based processors and at least one layer made up of non-quantum-based processors. This approach allows for extremely secure data transfer with high data compression.

Abstract: Spiking events in a spiking neural network may be processed via a memory system. A memory system may store data corresponding to a group of destination neurons. The memory system may, at each time interval of a SNN, pass through data corresponding to a group of pre-synaptic spike events from respective source neurons. The data corresponding to the group of pre-synaptic spike events may be subsequently stored in the memory system.

Abstract: Methods, systems, and non-transitory computer readable media are disclosed for accurately and efficiently generating groups of images portraying semantically similar objects for utilization in building machine learning models. In particular, the disclosed system utilizes metadata and spatial statistics to extract semantically similar objects from a repository of digital images. In some embodiments, the disclosed system generates color embeddings and content embeddings for the identified objects. The disclosed system can further group similar objects together within a query space by utilizing a clustering algorithm to create object clusters and then refining and combining the object clusters within the query space. In some embodiments, the disclosed system utilizes one or more of the object clusters to build a machine learning model.

Abstract: A neuron circuit, which electronically applies the working principle of the neurons in human brain, controls an input signal according to a set threshold value, and allows to provide an output signal above the threshold value. The neuron circuit controls an input signal according to a set threshold value and allows for an output signal above the threshold value, for determining the size of the threshold value of the circuit, and has at least one threshold resistor, at least one bias resistor, at least one decaying resistor, and at least one switching unit connected to at least one of these resistors.

Abstract: Systems and methods for extracting rule lists from tree ensembles are provided. A system extracts first stage candidate rules from individual trees. The system identifies the first stage candidate rules that satisfy a precision threshold and places those rules in a solution set. Subsequently, a determination is made whether a further stage is needed based on whether a predetermined number of positive data samples of the data set are covered by the solution set. In the further stage, the system generates next stage candidate rules from previous stage candidate rules that have not been pruned and identifies the next stage candidate rules that satisfy the precision threshold, placing those rules in the solution set. A simplified rule list is generated by identifying a minimum subset of rules in the solution set that covers the positive data samples within the precision threshold.

Abstract: Disclosed is a synaptic circuit including a weight memory that stores a weight value, a current-mode digital-to-analog converter (C-DAC) circuit that receives the weight value from the weight memory and supplies a current based on the weight value, a parasitic capacitor correction circuit that receives the weight value from the weight memory and to correct a value of parasitic capacitance generated by the C-DAC circuit based on the weight value, and a pre-discharge circuit that drains charges accumulated by the parasitic capacitance.

Abstract: A neuromorphic computing method includes comparing a maximum number of axons in which a size of a crossbar of a hardware-based node is considered with a number of input neurons, when the number of input neurons exceeds the maximum number of axons, grouping some of input neurons in consideration of the maximum number of axons, obtaining a spike output for a generated group, and inputting the spike output, together with remaining input neurons that are not included in the group, to any one node, and then processing the spike output.

Abstract: Sensor logs corresponding to a first machine are accessed. Each sensor log spans at least a first period. First computer readable logs corresponding to the first machine are accessed. Each computer readable log spans at least the first period, the computer readable logs include a maintenance log including maintenance task objects, each maintenance task object includes a time and a maintenance task type. A set of statistical metrics are derived from the sensor logs. A set of log metrics are derived from the computer readable logs. Using a risk model that receives the statistical metrics and log metrics as inputs, fault probabilities or risk scores indicative of one or more fault types occurring in the first machine within a second period are determined.

Abstract: Systems, devices, and methods related to safety monitoring in a factory using an artificial neural network are described. For example, the system can use a plurality of sensors installed at different locations of a factory to generate a plurality of streams of sensor data. At least one memory device can be configured in the system to perform matrix computations of the artificial neural network according to the plurality of streams of sensor data written into at least one memory device. Based on an output of the artificial neural network responsive to the plurality of streams of sensor data, the system generates an event identification representative of a hazard or anomaly in the factory and activates safety control or notification responsive to the event identification.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing a sequence-to-sequence model that is recurrent in depth while employing self-attention to combine information from different parts of sequences.

Abstract: Disclosed is operation method of an encoder that receives a continuous time-series signal and respectively transmits first to N-th input signals to first to N-th input neuron circuits of spike neural network circuit. The method of operating the encoder includes receiving the continuous time-series signal, generating a plurality of discrete quantum signals by sampling and quantizing the continuous time-series signal, selecting first to N-th discrete quantum signals among the plurality of discrete quantum signals, matching the selected first to N-th discrete quantum signals with the first to N-th input neuron circuits, respectively, identifying discrete quantum signals, each of which has a quantum level different from a quantum level of a previous discrete quantum signal, from among the second to N-th discrete quantum signals, and activating the input signals to be transmitted to the input neuron circuits corresponding to the identified discrete quantum signals and the first discrete quantum signal.

Abstract: Various embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for performing predictive structural analysis. Certain embodiments of the present invention utilize systems, methods, and computer program products that perform predictive structural analysis using at least one of techniques using time bound code transition likelihood data objects, techniques using cross-code relationship values, techniques using augmented entity-code occurrence data objects, techniques using per-pathway text representations of inferred occurrence pathways of a one or more individual historic code occurrences, techniques using polygenic risk score (PRS) measures, and/or the like.

Abstract: A method, a computing device, and a non-transitory machine-readable medium for classifying documents. A document collection is sorted into a plurality of categories. A classifier corresponding to a category of the plurality of categories is trained to output a probability that a document associated with the category is of a selected type (e.g., confidential). The training includes determining, by the processor, that a cardinality of a set of negative samples in a train set is not above a pipeline threshold but is at least one and training the classifier via a first pipeline and a second pipeline using a training group that includes a first portion of a group of positive samples in the train set, a second portion of a set of negative samples in the train set, and a third portion of a group of unlabeled samples in the train set.

Abstract: Embodiments of the present disclosure provide a stochastic neural network with training where the inputs to the gating function are not dependent on the control values of the previous layers' learnable weights but rather from a set of control-information weights, which may be fixed or adjustable. In various embodiments, a control node contains one or more control-parameters which control a set of activations and one or more information-parameters for the control node's output, to be consumed as inputs to the subsequent layers of control nodes. In various embodiments, one or more control-parameters and one or more information-parameters may be shared.

Abstract: In an example embodiment, a technique is presented that accesses training data that includes information about items, queries for items, and labels for the combinations of items and queries. The labels may correspond to different events, and there may be multiple different labels for the same combination of item and query. A machine learned model is then trained to learn a function for embedding each item to which a label pertains and a function for embedding each query to which a label pertains. Then, for each item in the training data, the items are embedded using the machine learned model, and the item embeddings for the item are concatenated into a single item embedding. At inference time, a similar concatenation is performed for multiple query embeddings. The concatenated embeddings are then used as input to an approximate k-nearest neighbor search function.

Abstract: Some embodiments provide a neural network inference circuit for executing a neural network that includes multiple nodes that use state data from previous executions of the neural network. The neural network inference circuit includes (i) a set of computation circuits configured to execute the nodes of the neural network and (ii) a set of memories configured to implement a set of one or more registers to store, while executing the neural network for a particular input, state data generated during at least two executions of the network for previous inputs. The state data is for use by the set of computation circuits when executing a set of the nodes of the neural network for the particular input.