Abstract: Video processing with a multi-quality loop filter using a multi-task neural network is performed by at least one processor and includes generating a first set of masked weight parameters, based on an input and a plurality of quantization parameter values with a corresponding first set of masks and first plurality of weight parameters, for a first set of shared neural network layers, selecting a second set of task specific neural network layers for the plurality of quantization parameter values with a second plurality of weight parameters, based on the plurality of quantization parameter values, computing an inference output, based on the first set of masked weight parameters and the second plurality of weight parameters, and outputting the computed inference output as an enhanced result.

Abstract: Disclosed is a hyperspectral data analysis method based on a semi-supervised learning strategy, which includes: hyperspectral sample data is acquired; a sample training set and a prediction set are constructed, herein an unlabeled prediction set sample is used; a regression network based on a generative adversarial network is constructed, including a generator network that generates a sample, and a discriminator/regressor network that has functions of judging the authenticity of the sample and outputting a quantitative analysis value at the same time; a loss function of the generative adversarial network is constructed, including a loss function of the discriminator, a loss function of the regressor, and a loss function of the generator with a sample distribution matching function. The generative adversarial network is used to generate a sample. A sample distribution matching strategy is used to supplement an existing unlabeled sample set. So, the accuracy of hyperspectral quantitative analysis is improved.

Abstract: The present invention discloses a spiking neural network for classifying input signals. The spiking neural network comprises a plurality of spiking neurons, and a plurality of synaptic elements interconnecting the spiking neurons to form the network. Each synaptic element is adapted to receive a synaptic input signal and apply a weight to the synaptic input signal to generate a synaptic output signal, the synaptic elements being configurable to adjust the weight applied by each synaptic element. Furthermore, each of the spiking neurons is adapted to receive one or more of the synaptic output signals from one or more of the synaptic elements, and generate a spatio-temporal spike train output signal in response to the received one or more synaptic output signals. The spiking neural network is partitioned into multiple sub-networks, wherein each sub-network comprises a sub-set of the spiking neurons connected to receive synaptic output signals from a sub-set of the synaptic elements.

Abstract: The present application provides a model gradient update method and device, for use in improving the accuracy of model training. A central server repeatedly executes a gradient update process until a stop condition is satisfied. One gradient update process comprises: receiving first gradients respectively sent by multiple nodes, the first gradients being obtained by each node using sample data to train a model to be trained of the node one or more times; obtaining a second gradient one the basis of the multiple first gradients and the probability of each node in the present gradient update process, the probability of each node in the present gradient update process being determined by an Actor-Critic network one the basis of the probability of each node in the last gradient update process; and sending the second gradient to the multiple nodes, respectively.

Abstract: A GIS mechanical fault diagnosis method and the device are disclosed. The method includes: collecting vibration signals to be measured of various excitation sources of GIS in mechanical operation; performing wavelet packet-feature entropy vector extraction on the vibration signals to be measured, when it is determined that the vibration signals to be measured are abnormal according to standard vibration signals in the normal state; inputting the extracted wavelet packet-feature entropy vectors into the pre-trained BP neural network for GIS mechanical fault identification, and outputting the corresponding fault. The disclosure integrates the vibration signals under the action of various excitation sources, extracts the feature entropy vectors according to the entropy theory, and constructs and trains a BP neural network that can classify and recognize various GIS mechanical faults, so as to perform comprehensive and effective GIS mechanical faults diagnose.

Abstract: A query-based generic end-to-end molecular optimization (“QMO”) system framework, method and computer program product for optimizing molecules, such as for accelerating drug discovery. The QMO framework decouples representation learning and guided search and applies to any plug-in encoder-decoder with continuous latent representations. QMO framework directly incorporates evaluations based on chemical modeling, analysis packages, and pre-trained machine-learned prediction models for efficient molecule optimization using a query-based guided search method based on zeroth order optimization. The QMO features efficient guided search with molecular property evaluations and constraints obtained using the predictive models and chemical modeling and analysis packages.

Abstract: Systems and methods are described for training a model for detecting manufacturing anomalies. A test response parameter is identified at a computing device, and a first plurality of component waveforms associated with the test response parameter are received at the computing device. Each waveform of the plurality of waveforms comprises a plurality of datapoints. A model is generated at the computing device, and the model is trained at the computing device and on the first plurality of component waveforms, thereby generating one or more parameters associated with the model. A second plurality of component waveforms associated with the test response parameter is received, and the trained model is accessed. It is indicated using the trained model, whether any of the second plurality of component waveforms comprises an anomaly. For each indicated waveform, the indicated waveform is reviewed and, for each reviewed waveform not comprising an anomaly, the waveform is labelled.

Abstract: Computer systems and computer-implemented methods train a neural network, by: (a) computing for each datum in a set of training data, activation values for nodes in the neural network and estimates of partial derivatives of an objective function for the neural network for the nodes in the neural network; (b) selecting a target node of the neural network and/or a target datum in the set of training data; (c) selecting a target-specific improvement model for the neural network, wherein the target-specific improvement model, when added to the neural network, improves performance of the neural network for the target node and/or the target datum, as the case may be; (d) training the target-specific improvement model; (e) merging the target-specific improvement model with the neural network to form an expanded neural network; and (f) training the expanded neural network.

Abstract: A method is provided that includes accessing a multivariate time series of flight data for an aircraft, and iteratively performing runs of genetic programming on groups of the sensors. A population of computer programs is randomly generated from a selected group of the plurality of sensors, and primitive functions selected from a library of primitive functions. The population is iteratively transformed into new generations of the population, and includes sub-rankings of the group of sensors based on a quantitative fitness determined according to selected fitness criterion. A ranking of the group of sensors from the sub-rankings of the group of sensors is produced. An aggregate ranking of the plurality of sensors is produced from the ranking of the group of sensors over a plurality of iterations. And the subset of sensors is selected from the aggregate ranking of the plurality of sensors, and according to selected optimization criterion.

Abstract: A method for driving path prediction is provided. The method concatenates past trajectory features and lane centerline features in a channel dimension at an agent's respective location in a top view map to obtain concatenated features thereat. The method obtains convolutional features derived from the top view map, the concatenated features, and a single representation of the training scene the vehicle and agent interactions. The method extracts hypercolumn descriptor vectors which include the convolutional features from the agent's respective location in the top view map. The method obtains primary and auxiliary trajectory predictions from the hypercolumn descriptor vectors. The method generates a respective score for each of the primary and auxiliary trajectory predictions.

Abstract: Techniques for optimizing and deploying deep neural network (CNN) machine learning models for inference using static analysis are described. A method includes obtaining a deep neural network (DNN) machine learning (ML) model, generating an intermediate representation for the ML model, the intermediate representation including one or more nodes corresponding to one or more operators utilized by the ML model, identifying, for at least one node of the intermediate representation, an optimized schedule for at least one operator corresponding to the at least one node using a static analysis that is based on a hardware-specific cost model, generating an optimized intermediate representation using the optimized schedule that is optimized for execution on a hardware platform, and generating code corresponding to the ML model based at least in part on the optimized intermediate representation, wherein the code is specific to the hardware platform.

Abstract: A provider network implements a machine learning deployment service for generating and deploying packages to implement machine learning at connected devices. The service may receive from a client an indication of an inference application, a machine learning framework to be used by the inference application, a machine learning model to be used by the inference application, and an edge device to run the inference application. The service may then generate a package based on the inference application, the machine learning framework, the machine learning model, and a hardware platform of the edge device. To generate the package, the service may optimize the model based on the hardware platform of the edge device and/or the machine learning framework. The service may then deploy the package to the edge device. The edge device then installs the inference application and performs actions based on inference data generated by the machine learning model.

Abstract: Forecasting resource allocation is disclosed. An example method includes receiving operating data from a resource; applying periodicity tests to the received operating data using a plurality of sketches of time series of prior operating data, the periodicity tests generating periodic components; applying regression models to the received operating data, the regression models collectively generating a trend component, each regression model being applied over a different time scale of a plurality of time scales; computing a trend model using the periodic components and a trend component; determining a random process describing the historical evolution of the trend model; and calculating and providing a mean prediction, an upper bound, and a lower bound for resource utilization at a future time using the trend model and a predicted distribution.

Abstract: Methods and apparatus for quantum assisted optimization. In one aspect, a method includes obtaining a set of initial input states, applying one or more of (i) dynamical thermal fluctuations and (ii) cluster update algorithms to the set of input states and subsequent input states when the states evolve within the classical information processors, applying dynamical quantum fluctuations to the set of input states and subsequent states when the states evolve within the quantum systems and repeating the application steps until a desirable output state is obtained.

Abstract: Provided is a process including: receiving a data token to be passed from a first node to a second node; retrieving machine learning model attributes from a collection of one or more of the sub-models of a federated machine-learning model; determining based on the machine learning model attributes, that the data token is learning relevant to members of the collection of one or more of the sub-models and, in response, adding the data toke to a training set to be used by at least some members of the collection of one or more of the sub-models; determining a collection of data tokens to transmit from the second node to a third node of the set of nodes participating in a federated machine-learning model; and transmitting the collection of data tokens.

Abstract: A parallel processing method and apparatus for a neural network model. The parallel processing method includes extracting metadata of a target layer included in a target model, measuring a similarity between the target layer and each of reference layers by comparing the metadata of the target layer to reference metadata of each of the reference layers, selecting a corresponding layer among the reference layers based on the similarities, and generating a parallelization strategy for the target layer based on a reference parallelization strategy matching the corresponding layer.

Abstract: Described are a system, method, and computer program product for dynamic node classification in temporal-based machine learning classification models. The method includes receiving graph data of a discrete time dynamic graph including graph snapshots, and node classifications associated with all nodes in the discrete time dynamic graph. The method includes converting the discrete time dynamic graph to a time-augmented spatio-temporal graph and generating an adjacency matrix based on a temporal walk of the time-augmented spatio-temporal graph. The method includes generating an adaptive information transition matrix based on the adjacency matrix and determining feature vectors based on the nodes and the node attribute matrix of each graph snapshot.

Abstract: A method and system for identifying entangled photons includes generating a plurality of sets of four entangled photons, wherein one pair of photons of each set are time correlated, thereby indicating that another pair of four entangled photons are entangled. Quantum metadata comprising a time window corresponding to the generated plurality of sets of four entangled photons is collected. A coincidence of one pair of photons of each of the plurality of the sets of four entangled photons is determined. A state value of at least one photon of the other pair of each of the number of the sets of four entangled photons is determined. Ordered lists of coincidences are compared to ordered lists of state values to determine entangled state information. Time window are compared to times corresponding to the ordered lists. Error conditions are generated if conditions are met.

Abstract: Computer systems and computer-implemented methods train a neural network iteratively training, through machine learning. The iterative training comprises imposing a first is-not-equal-to regularization link between first and second nodes, where imposing the first is-not-equal-to regularization link between the two nodes comprises adding, during back-propagation of partial derivatives through the neural network for a datum in a training data set, a first regularization cost to a network error loss function for the first node that is inversely proportional to a difference between an activation value for the first node for the datum and an activation value for the second node for the datum.

Abstract: Provided are systems, methods, and computer program products for generating node embeddings. The system includes at least one processor programmed or configured to generate a graph comprising a plurality of nodes, generate an embedding for each node of the plurality of nodes, each embedding comprising at least one polar angle and a vector length, store each embedding of a plurality of embeddings in memory, and in response to processing the graph with a machine-learning algorithm, convert at least one embedding of the plurality of embeddings to Cartesian coordinates.