Abstract: In some embodiments, an archipelago model is provided for distributed execution of evolutionary computing techniques. In some embodiments, an archipelago manager computing device manages a centralized archipelago management queue, and provides population subsets to a plurality of island computing devices. The archipelago manager computing device receives candidate solutions from the island computing devices, stores the candidate solutions in the archipelago management queue, and transmits candidate solutions from the archipelago management queue to the island computing devices in order to exchange candidate solutions between the island computing devices. The use of an archipelago management queue allows transfer of candidate solutions between different island computing devices, is robust to failure of any given island computing device, and does not require homogeneity within the plurality of island computing devices.

Abstract: An apparatus includes an integrated circuit (IC), which includes complementary metal oxide semiconductor (CMOS) circuitry. The CMOS circuitry includes a p-channel transistor network that includes at least one p-channel transistor having a gate-induced drain leakage (GIDL) current. The IC further includes a native metal oxide semiconductor (MOS) transistor coupled to supply a bias voltage to the at least one p-channel transistor to reduce the GIDL current of the at least one p-channel transistor.

Abstract: A system including a confidence assessment module that implements a neural network to assess the likelihood that codes associated with a patient's encounter with a healthcare organization are accurate. The confidence assessment module may also be incrementally trained.

Abstract: A technique includes modifying a neural network model to sparsify the model. The model includes a plurality of kernel element weights, which are parameterized according to a plurality of dimensions. Modifying the model includes, in a given iteration of the plurality of iterations, training the model based on a structure regularization in which kernel element weights that share a dimension in common are removed as a group to create corresponding zero kernel elements in the model; and compressing the model to exclude zero kernel element weights from the model to prepare the model to be trained in another iteration.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for generating a target distribution schedule for providing electronic communications based on predicted behavior rates by utilizing a genetic algorithm and one or more objective functions. For example, the disclosed systems can generate predicted behavior rates by training and utilizing one or more behavior prediction models. Based on the predicted behavior rates, the disclosed systems can further utilize a genetic algorithm to apply objective functions to generate one or more candidate distribution schedules. In accordance with the genetic algorithm, the disclosed systems can select a target distribution schedule for a particular user/client device. The disclosed systems can thus provide one or more electronic communications to individual users based on respective target distribution schedules.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a student neural network. In one aspect, there is provided a method comprising: processing a training input using the student neural network to generate a student neural network output comprising a respective score for each of a plurality of classes; processing the training input using a brain emulation neural network to generate a brain emulation neural network output comprising a respective score for each of the plurality of classes; and adjusting current values of the student neural network parameters using gradients of an objective function that characterizes a similarity between: (i) the student neural network output for the training input, and (ii) the brain emulation neural network output for the training input.

Abstract: Some embodiments provide a compiler for optimizing the implementation of a machine-trained network (e.g., a neural network) on an integrated circuit (IC). In some embodiments, the compiler determines whether sparsity requirements of channels implemented on individual cores are met on each core. If the sparsity requirement is not met, the compiler, in some embodiments, determines whether the channels of the filter can be rearranged to meet the sparsity requirements on each core and, based on the determination, either rearranges the filter channels or implements a solution to non-sparsity.

Abstract: In an artificial neural network, integrality refers to the degree to which a neuron generates, for a given set of inputs, outputs that are near the border of the output range of a neuron. From each neural network of a pool of trained neural networks, a group of neurons with a higher integrality is selected to form a neural network tunnel (“tunnel”). The tunnel must include all input neurons and output neurons from the neural network, and some of the hidden neurons. Tunnels generated from each neural network in a pool are merged to form another neural network. The new network may then be trained.

Abstract: A processor may monitor frequency data related to a user metric of a user during a measurement window. The user metric may relate to the user's use of a computer implemented environment. The processor may simplify the frequency data related to the user metric, resulting in a set of simplified frequency data. The processor may input the set of simplified frequency data into a neural network. The neural network may determine a likelihood of a negative user experience for the user. The processor may alter a parameter of the first user environment based on the likelihood.

Abstract: A method, system, and computer program product for resource management are described. The method includes selecting trouble regions within the service area, generating clustered regions, and training a trouble forecast model for the trouble regions for each type of damage, the training for each trouble region using training data from every trouble region within the clustered region associated with the trouble region. The method also includes applying the trouble forecast model for each trouble region within the service area for each type of damage, determining a trouble forecast for the service area for each type of damage based on the trouble forecast for each of the trouble regions within the service area, and determining a job forecast for the service area based on the trouble forecast for the service area, wherein the managing resources is based on the job forecast for the service area.

Abstract: Methods, apparatus, and processor-readable storage media for data integration demand management using artificial intelligence are provided herein. An example computer-implemented method includes obtaining at least one data integration demand, wherein the at least one data integration demand comprises textual information provided by at least one user; determining multiple parameters of the at least one data integration demand by applying one or more machine learning natural language processing techniques to at least a portion of the textual information provided by the at least one user; generating at least one delivery date prediction for the at least one data integration demand by applying one or more artificial intelligence techniques to the multiple determined parameters of the at least one data integration demand; and performing one or more automated actions based at least in part on the at least one generated delivery date prediction.

Abstract: A semiconductor device that is less influenced by variations in characteristics between transistors or variations in a load, and is efficient even for normally-on transistors is provided. The semiconductor device includes at least a transistor, two wirings, three switches, and two capacitors. A first switch controls conduction between a first wiring and each of a first electrode of a first capacitor and a first electrode of a second capacitor. A second electrode of the first capacitor is connected to a gate of the transistor. A second switch controls conduction between the gate and a second wiring. A second electrode of the second capacitor is connected to one of a source and a drain of the transistor. A third switch controls conduction between the one of the source and the drain and each of the first electrode of the first capacitor and the first electrode of the second capacitor.

Abstract: A computer-implemented method optimizing genetic algorithms for finding solutions to a provided problem is described. The method implements a multi-arm bandit algorithm to determine performance scores for candidate individuals from a candidate pool in dependence on successes and failures of the one or more candidates. The method evolves the candidate individuals in the candidate pool by performing evolution steps including: determining a fitness score for each of the candidate individuals in the candidate pool in dependence on the performance scores for the candidate individuals, discarding candidate individuals from the candidate pool in dependence upon their assigned performance measure, and adding, to the candidate pool, a new candidate individual procreated from candidate individuals remaining in the candidate pool after the discarding of the candidate individuals.

Abstract: A delay circuit includes precharge and discharge transistors configured to receive an input signal. The delay circuit also includes a resistor coupled to the precharge transistor having a negative temperature coefficient to thereby form a node. A capacitive device and an inverter are coupled to the node. The inverter produces an output signal. Responsive to the input signal having a first polarity, the precharge transistor is configured to be turned on and the discharge transistor is configured to be turned off to thereby cause current to flow through the precharge transistor to the capacitive device to thereby charge the capacitive device. Responsive to the input signal having a second polarity, the precharge and discharge transistors are configured to change state to thereby cause charge from the capacitive device to discharge through the resistor and through the discharge transistor. The voltage on the node decays to a level which eventually causes the inverter's output to change state.

Abstract: A nervous system emulator engine includes working computational models of the vertebrate nervous system to generate lifelike animal behavior in a robot. These models include functions representing several anatomical features of the vertebrate nervous system, such as spinal cord, brainstem, basal ganglia, thalamus and cortex. The emulator engine includes a hierarchy of controllers in which controllers at higher levels accomplish goals by continuously specifying desired goals for lower-level controllers. The lowest levels of the hierarchy reflect spinal cord circuits that control muscle tension and length. Moving up the hierarchy into the brainstem and midbrain/cortex, progressively more abstract perceptual variables are controlled. The nervous system emulator engine may be used to build a robot that generates the majority of animal behavior, including human behavior. The nervous system emulator engine may also be used to build working models of nervous system functions for clinical experimentation.

Abstract: A method includes: generating a plurality of individuals of a current generation in accordance with a plurality of individuals of a previous generation to acquire values of an objective function for individuals each representing a variable by evolutionary computation; calculating, for each of partial individuals of the plurality of individuals of the current generation generated by the generating processing, a first value of the objective function by a predetermined method; approximately calculating, for each of the plurality of individuals of the current generation, a second value of the objective function with lower precision than the predetermined method; computing a fitness difference representing a difference between the plurality of individuals of the current generation in accordance with the first value or the second value; and controlling the precision of the approximate calculation based on the fitness difference and a precision difference between the first value and the second value.

Abstract: A technology is provided for predicting congestion or crowding of services over a future time interval, and may be utilized for forecasting congestion in a hospital emergency department. One embodiment of this technology comprises a decision support tool for resources management to prevent overcrowding and long waiting times, or for mitigating ED congestion by, for example, warning hospital managers that a significant likelihood exists of ED congestion over a future time frame, or automatically initiating mitigative actions. A time series of consecutive ED arrivals timestamps is processed to determine a presence (or absence) of positive autocorrelation or self-similarity and estimate Hurst exponent values to generate a forecast model. The forecast model is utilized to determine future ED demand.

Abstract: Systems and methods provide a learned difference metric that operates in a wide artifact space. An example method includes initializing a committee of deep neural networks with labeled distortion pairs, iteratively actively learning a difference metric using the committee and psychophysics tasks for informative distortion pairs, and using the difference metric as an objective function in a machine-learned digital file processing task. Iteratively actively learning the difference metric can include providing an unlabeled distortion pair as input to each of the deep neural networks in the committee, a distortion pair being a base image and a distorted image resulting from application of an artifact applied to the base image, obtaining a plurality of difference metric scores for the unlabeled distortion pair from the deep neural networks, and identifying the unlabeled distortion pair as an informative distortion pair when the difference metric scores satisfy a diversity metric.

Abstract: The present invention addresses the problem of implementing a neural network using a small-scale circuit by simplifying the multiplication of the input data by weight data. The neural network circuit according to the present invention is configured from: a means for multiplying input data by a rounded value of the mantissa part of weight data; a means for shifting the multiplication result by the number of bits of the rounded value; a means for adding the shifted result to the original input data; and a means for shifting the addition result by the number of bits of the exponent part of the weight.

Abstract: A separate quantization method of forming a combination of 4-bit and 8-bit data of a neural network is disclosed. When a training data set and a validation data set exist, a calibration manner is used to determine a threshold for activations of each of a plurality of layers of a neural network model, so as to determine how many of the activations to perform 8-bit quantization. In a process of weight quantization, the weights of each layer are allocated to 4-bit weights and 8-bit weights according to a predetermined ratio, so as to make the neural network model have a reduced size and a combination of 4-bit and 8-bit weights.