Abstract: The present disclosure generally relates to a primary load management system configured to execute machine learning and artificial intelligence techniques to generate predictions of access-right requests that are or are likely to be invalid before the access-right requests are processed for assignment to users or user devices. The present disclosure relates to systems and methods that collect a data set representing characteristics of user devices as the user devices interact with various systems of the primary load management system and train a machine-learning model to predict invalid access-right requests using the collected data set. The collected data set may include a log line that represents each user device, and each log line may be labeled based on an invalidity evaluation. New access-right requests can be processed using the trained machine-learning model to determine whether or not to assign access rights in response to the access-right request.

Abstract: In an approach for compressing a neural network, a processor receives a neural network, wherein the neural network has been trained on a set of training data. A processor receives a compression ratio. A processor compresses the neural network based on the compression ratio using an optimization model to solve for sparse weights. A processor re-trains the compressed neural network with the sparse weights. A processor outputs the re-trained neural network.

Abstract: Disclosed is an electronic apparatus. The electronic apparatus may include a memory configured to store one or more training data generation models and an artificial intelligence model, and a processor configured to generate personal training data that reflects a characteristic of a user using the one or more training data generation models, train the artificial intelligence model using the personal learning data as training data, and store the trained artificial intelligence model in the memory.

Abstract: A method in a hardware implementation of a Convolutional Neural Network (CNN), includes receiving a first subset of data having at least a portion of weight data and at least a portion of input data for a CNN layer and performing, using at least one convolution engine, a convolution of the first subset of data to generate a first partial result; receiving a second subset of data comprising at least a portion of weight data and at least a portion of input data for the CNN layer and performing, using the at least one convolution engine, a convolution of the second subset of data to generate a second partial result; and combining the first partial result and the second partial result to generate at least a portion of convolved data for a layer of the CNN.

Abstract: A pure integer quantization method for a lightweight neural network (LNN) is provided. The method includes the following steps: acquiring a maximum value of each pixel in each of the channels of the feature map of a current layer; dividing a value of each pixel in each of the channels of the feature map by a t-th power of the maximum value, t?[0,1]; multiplying a weight in each of the channels by the maximum value of each pixel in each of the channels of the corresponding feature map; and convolving the processed feature map with the processed weight to acquire the feature map of a next layer. The algorithm is verified on SkyNet and MobileNet respectively, and lossless INT8 quantization on SkyNet and maximum quantization accuracy so far on MobileNetv2 are achieved.

Abstract: Systems and methods for controlling a material extrusion device to extrude a filament of an ink are provided. An extrusion printing control system collects from one or more sensors measurements representing an internal state of material extrusion processing during extrusion of the filament. In addition, the system collects an image of the filament as the filament is extruded. The system applies a classifier to the collected image to generate an image-derived state characterizing the filament. Based on the internal state and the image-derived state, the system estimates a derived state using a model. The system determines control parameters using the model to achieve a desired quality of the filament by minimizing a cost function based on the internal state, the image-derived state, the derived state, and constraints of the material extrusion device. Finally, the system provides the control parameters to a controller of the material extrusion device.

Abstract: Embodiments of the invention provide a system, media, and method for deep learning applications in physical design verification. Generally, the approach includes maintaining a pattern library for use in training machine learning model(s). The pattern library being generated adaptively and supplemented with new patterns after review of new patterns. In some embodiments, multiple types of information may be included in the pattern library, including validation data, and parameter and anchoring data used to generate the patterns. In some embodiments, the machine learning processes are combined with traditional design rule analysis. The patterns being generated and adapted using a lossless process that encodes the information of a corresponding area of a circuit layout.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for selecting actions from large discrete action sets. One of the methods includes receiving a particular observation representing a particular state of an environment; and selecting an action from a discrete set of actions to be performed by an agent interacting with the environment, comprising: processing the particular observation using an actor policy network to generate an ideal point; determining, from the points that represent actions in the set, the k nearest points to the ideal point; for each nearest point of the k nearest points: processing the nearest point and the particular observation using a Q network to generate a respective Q value for the action represented by the nearest point; and selecting the action to be performed by the agent from the k actions represented by the k nearest points based on the Q values.

Abstract: A method for building an artificial intelligence (AI) model. The method includes accessing data related to monitored behavior of a user. The data is classified, wherein the classes include an objective data class identifying data relevant to a group of users including the user, and a subjective data class identifying data that is specific to the user. Objective data is accessed and relates to monitored behavior of a plurality of users including the user. The method includes providing as a first set of inputs into a deep learning engine performing AI the objective data and the subjective data of the user, and a plurality of objective data of the plurality of users. The method includes determining a plurality of learned patterns predicting user behavior when responding to the first set of inputs. The method includes building a local AI model of the user including the plurality of learned patterns.

Abstract: An accelerator for processing of a convolutional neural network (CNN) includes a compute core having a plurality of compute units. Each compute unit includes a first memory cache configured to store at least one vector in a map trace, a second memory cache configured to store at least one vector in a kernel trace, and a plurality of vector multiply-accumulate units (vMACs) connected to the first and second memory caches. Each vMAC includes a plurality of multiply-accumulate units (MACs). Each MAC includes a multiplier unit configured to multiply a first word that of the at least one vector in the map trace by a second word of the at least one vector in the kernel trace to produce an intermediate product, and an adder unit that adds the intermediate product to a third word to generate a sum of the intermediate product and the third word.

Abstract: Techniques for sequence to sequence (S2S) model building and/or optimization are described. For example, a method of receiving a request to build a sequence to sequence (S2S) model for a use case, wherein the request includes at least a training data set, generating parts of a S2S algorithm based on the at least one use case, determined parameters, and determined hyperparameters, and training a S2S algorithm built from the parts of the S2S algorithm using the training data set to generate the S2S model is detailed.

Abstract: Neural Architecture Search (NAS) is a laborious process. Prior work on automated NAS targets mainly on improving accuracy but lacked consideration of computational resource use. Presented herein are embodiments of a Resource-Efficient Neural Architect (RENA), an efficient resource-constrained NAS using reinforcement learning with network embedding. RENA embodiments use a policy network to process the network embeddings to generate new configurations. Example demonstrates of RENA embodiments on image recognition and keyword spotting (KWS) problems are also presented herein. RENA embodiments can find novel architectures that achieve high performance even with tight resource constraints. For the CIFAR10 dataset, the tested embodiment achieved 2.95% test error when compute intensity is greater than 100 FLOPs/byte, and 3.87% test error when model size was less than 3M parameters.

Abstract: In many industrial settings, a process is repeated many times, for instance to transform physical inputs into physical outputs. To detect a situation involving such a process in which errors are likely to occur, information about the process may be collected to determine time-varying feature vectors. Then, a drift value may be determined by comparing feature vectors corresponding with different time periods. When the drift value crosses a designated drift threshold, a predicted outcome value may be determined by applying a prediction model. Sensitivity values may be determined for different features, and elements of the process may then be updated based at least in part on the sensitivity values.

Abstract: In various embodiments, a prediction subsystem automatically predicts a level of network availability of a device network. The prediction subsystem computes a set of predicted attribute values for a set of devices attributes associated with the device network based on a trained recurrent neural network (RNN) and set(s) of past attribute values for the set of device attributes. The prediction subsystem then performs classification operation(s) based on the set of predicted attribute values and one or more machine-learned classification criteria. The result of the classification operation(s) is a network availability data point that predicts a level of network availability of the device network. Preemptive action(s) are subsequently performed on the device network based on the network availability data point. By performing the preemptive action(s), the amount of time during which network availability is below a given level can be substantially reduced compared to prior art, reactive approaches.

Abstract: Techniques regarding providing artificial intelligence problem descriptions are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory, and that can execute the computer executable components stored in the memory. The computer executable components can include, at least: a query component that generates key performance indicators from a query, determines a subset of key performance indicators that individually have a performance below a threshold, and maps the subset of key performance indicators to operational metrics; a learning component that generates, using artificial intelligence, problem descriptions from one or more of the subset of key performance indicators or the operational metrics and transmits the problem descriptions to a database.

Abstract: Systems, apparatuses and methods may provide for technology that adjusts a plurality of weights in a neural network model and adjusts a plurality of activation functions in the neural network model. The technology may also output the neural network model in response to one or more conditions being satisfied by the plurality of weights and the plurality of activation functions. In one example, two or more of the activation functions are different from one another and the activation functions are adjusted on a per neuron basis.

Abstract: An optimization apparatus calculates a first portion, among energy change caused by change in value of a neuron of a neuron group, caused by influence of another neuron of the neuron group, determines whether to allow updating the value, based on a sum of the first and second portions of the energy change, and repeats a process of updating or maintaining the value according to the determination. An arithmetic processing apparatus calculates the second portion caused by influence of a neuron not belonging to the neuron group and an initial value of the sum. A control apparatus transmits data for calculating the second portion and the initial value to the arithmetic processing apparatus, and the initial value and data for calculating the first portion to the optimization apparatus, and receives the initial value from the arithmetic processing apparatus, and a value of the neuron group from the optimization apparatus.

Abstract: A digital circuit for accelerating computations of an artificial neural network model includes a pairs selection unit that selects different subsets of pairs of input vector values and corresponding weight vector values to be processed simultaneously at each time step; a sorting unit that simultaneously processes a vector of input-weight pairs wherein pair values whose estimated product is small are routed with a high probability to small multipliers, and pair values whose estimated product is greater are routed with a high probability to large multipliers that support larger input and output values; and a core unit that includes a plurality of multiplier units and a plurality of adder units that accumulate output results of the plurality of multiplier units into one or more output values that are stored back into the memory, where the plurality of multiplier units include the small multipliers and the large multipliers.

Abstract: A computer-implemented method includes instantiating a neural network including a recurrent cell. The recurrent cell includes a probabilistic state component. The method further includes training the neural network with a sequence of data. In an embodiment, the method includes extracting a deterministic finite automaton from the trained recurrent neural network and classifying a sequence with the extracted automaton.

Abstract: A mechanism is provided in a data processing system having a processor and a memory. The memory comprises instructions which are executed by the processor to cause the processor to implement a training system for finding an optimal surface for hierarchical classification task on an ontology. The training system receives a training data set and a hierarchical classification ontology data structure. The training system generates a neural network architecture based on the training data set and the hierarchical classification ontology data structure. The neural network architecture comprises an indicative layer, a parent tier (PT) output and a lower leaf tier (LLT) output. The training system trains the neural network architecture to classify the training data set to leaf nodes at the LLT output and parent nodes at the PT output. The indicative layer in the neural network architecture determines a surface that passes through each path from a root to a leaf node in the hierarchical ontology data structure.