Abstract: A treatment planning prediction method to predict a Dose-Volume Histogram (DVH) or Dose Distribution (DD) for patient data using a machine-learning computer framework is provided with the key inclusion of a Planning Target Volume (PTV) only treatment plan in the framework. A dosimetric parameter is used as an additional parameter to the framework and which is obtained from a prediction of the PTV-only treatment plan. The method outputs a Dose-Volume Histogram and/or a Dose Distribution for the patient including the prediction of the PTV-only treatment plan. The method alleviates the complicated process of quantifying anatomical features and harnesses directly the inherent correlation between the PTV-only plan and the clinical plan in the dose domain. The method provides a more robust and efficient solution to the important DVHs prediction problem in treatment planning and plan quality assurance.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for selecting a neural network to perform a particular machine learning task while satisfying a set of constraints.

Abstract: Systems and methods for generating embeddings for nodes of a corpus graph are presented. The embeddings correspond to aggregated embedding vectors for nodes of the corpus graph. Without processing the entire corpus graph to generate all aggregated embedding vectors, a relevant neighborhood of nodes within the corpus graph are identified for a target node of the corpus graph. Based on embedding information of the target node's immediate neighbors, and also upon neighborhood embedding information from the target node's relevant neighborhood, an aggregated embedding vector can be generated for the target node that comprises both an embedding vector portion corresponding to the target node, as well as a neighborhood embedding vector portion, corresponding to embedding information of the relevant neighborhood of the target node. Utilizing both portions of the aggregated embedding vector leads to improved content recommendation to a user in response to a query.

Abstract: A system is provided for classifying an instruction sequence with a machine learning model. The system may include at least one processor and at least one memory. The memory may include program code that provides operations when executed by the at least one processor. The operations may include: processing an instruction sequence with a trained machine learning model configured to detect one or more interdependencies amongst a plurality of tokens in the instruction sequence and determine a classification for the instruction sequence based on the one or more interdependencies amongst the plurality of tokens; and providing, as an output, the classification of the instruction sequence. Related methods and articles of manufacture, including computer program products, are also provided.

Abstract: Various systems, devices, and methods for operating on a data sequence. A system includes a set of circuits that form an input layer to receive a data sequence; first hardware computing units to transform the data sequence, the first hardware computing units connected using a set of randomly selected weights, a first hardware computing unit to: receive an input from a second hardware computing unit, determine a weight of a connection between the first and second hardware computing units using an identifier of the second hardware computing unit and a fixed random weight generator, and operate on the input using the weight to determine a state of the first hardware computing unit; and second hardware computing units to operate on states of the first computing units to generate an output based on the data sequence.

Abstract: Aspects of data modification for neural networks are described herein. The aspects may include a data modifier configured to receive input data and weight values of a neural network. The data modifier may include an input data configured to modify the received input data and a weight modifier configured to modify the received weight values. The aspects may further include a computing unit configured to calculate one or more groups of output data based on the modified input data and the modifier weight values.

Abstract: An artificial neural network (ANN) system includes a processor, a virtual overflow detection circuit and a data format controller. The processor performs node operations with respect to a plurality of nodes included in each layer of an ANN to obtain a plurality of result values of the node operations and performs a quantization operation on the obtained plurality of result values based on a k-th fixed-point format for a current quantization of the each layer to obtain a plurality of quantization values. The virtual overflow detection circuit generates a virtual overflow information indicating a distribution of valid bit numbers of the obtained plurality of quantization values. The data format controller determines a (k+1)-th fixed-point format for a next quantization of the each layer based on the generated virtual overflow information. An overflow and/or an underflow are prevented efficiently by controlling the fixed-point format using the virtual overflow.

Abstract: A neuromorphic chip includes synaptic cells including respective resistive devices, axon lines, dendrite lines and switches. The synaptic cells are connected to the axon lines and dendrite lines to form a crossbar array. The axon lines are configured to receive input data and to supply the input data to the synaptic cells. The dendrite lines are configured to receive output data and to supply the output data via one or more respective output lines. A given one of the switches is configured to connect an input terminal to one or more input lines and to changeably connect its one or more output terminals to a given one or more axon lines.

Abstract: A computer-implemented method for performing computer vision with reduced computational cost and improved accuracy can include obtaining, by a computing system including one or more computing devices, input data comprising an input tensor having one or more dimensions, providing, by the computing system, the input data to a machine-learned convolutional attention network, the machine-learned convolutional attention network including two or more network stages, and, in response to providing the input data to the machine-learned convolutional attention network, receiving, by the computing system, a machine-learning prediction from the machine-learned convolutional attention network. The convolutional attention network can include at least one attention block, wherein the attention block includes a relative attention mechanism, the relative attention mechanism including the sum of a static convolution kernel with an adaptive attention matrix.

Abstract: An improved computer implemented method and corresponding systems and computer readable media for improving performance of a deep neural network are provided to mitigate effects related to catastrophic forgetting in neural network learning. In an embodiment, the method includes storing, in memory, logits of a set of samples from a previous set of tasks (D1); and maintaining classification information from the previous set of tasks by utilizing the logits for matching during training on a new set of tasks (D2).

Abstract: The present disclosure provides an integrated circuit chip device and a related product. The integrated circuit chip device includes: a primary processing circuit and a plurality of basic processing circuits. The primary processing circuit or at least one of the plurality of basic processing circuits includes the compression mapping circuits configured to perform compression on each data of a neural network operation. The technical solution provided by the present disclosure has the advantages of a small amount of computations and low power consumption.

Abstract: The present disclosure provides directed to new, more efficient neural network architectures. As one example, in some implementations, the neural network architectures of the present disclosure can include a linear bottleneck layer positioned structurally prior to and/or after one or more convolutional layers, such as, for example, one or more depthwise separable convolutional layers. As another example, in some implementations, the neural network architectures of the present disclosure can include one or more inverted residual blocks where the input and output of the inverted residual block are thin bottleneck layers, while an intermediate layer is an expanded representation. For example, the expanded representation can include one or more convolutional layers, such as, for example, one or more depthwise separable convolutional layers. A residual shortcut connection can exist between the thin bottleneck layers that play a role of an input and output of the inverted residual block.

Abstract: An artificial intelligence system includes a neural network layer including an arithmetic operation circuit that performs an arithmetic operation of a sigmoid function. The arithmetic operation circuit includes a first circuit configured to perform an exponent arithmetic operation using a Napier's constant e as a base and output a first calculation result when an exponent in the exponent arithmetic operation is a negative number, wherein an absolute value of the exponent is used in the exponent arithmetic operation, and a second circuit configured to subtract the first calculation result obtained by the first circuit from 1 and output the subtracted value.

Abstract: A method and system for managing a dataset. An artificial intelligence (AI) model is to be used on the dataset. A data mask describes a labeling status of the data items. A loop is repeated until patience parameters are satisfied. The loop comprises receiving trusted labels provided by trusted labelers; updating the data mask; from a labelled data items subset, training the AI model; cloning the trained AI model into a local AI model on processing nodes; creating and chunking a randomized unlabeled subset into data subsets for dispatching to the processing nodes; receiving an indication that predicted label answers have been inferred by the processing nodes using the local AI model; computing a model uncertainty measurement from statistical analysis of the predicted label answers. The patience parameters include one or more of a threshold value on the model uncertainty measurement and information gain between different training cycles.

Abstract: A processor implemented method to update a sentence generation model includes: generating a target sentence corresponding to a source sentence using a first decoding model; calculating reward information associated with the target sentence using a second decoding model configured to generate a sentence in an order different from an order of the sentence generated by the first decoding model; and generating an updated sentence generation model by resetting a weight of respective nodes in the first decoding model based on the calculated reward information.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network used to select actions to be performed by an agent interacting with an environment. One of the methods includes obtaining data identifying (i) a first observation characterizing a first state of the environment, (ii) an action performed by the agent in response to the first observation, and (iii) an actual reward received resulting from the agent performing the action in response to the first observation; determining a pseudo-count for the first observation; determining an exploration reward bonus that incentivizes the agent to explore the environment from the pseudo-count for the first observation; generating a combined reward from the actual reward and the exploration reward bonus; and adjusting current values of the parameters of the neural network using the combined reward.

Abstract: A method includes maintaining respective episodic memory data for each of multiple actions; receiving a current observation characterizing a current state of an environment being interacted with by an agent; processing the current observation using an embedding neural network in accordance with current values of parameters of the embedding neural network to generate a current key embedding for the current observation; for each action of the plurality of actions: determining the p nearest key embeddings in the episodic memory data for the action to the current key embedding according to a distance measure, and determining a Q value for the action from the return estimates mapped to by the p nearest key embeddings in the episodic memory data for the action; and selecting, using the Q values for the actions, an action from the multiple actions as the action to be performed by the agent.

Abstract: A method for time series analysis of time-oriented usage data pertaining to computing resources of a computing system. A method embodiment commences upon collecting time series datasets, individual ones of the time series datasets comprising time-oriented usage data of a respective individual computing resource. A plurality of prediction models are trained using portions of time-oriented data. The trained models are evaluated to determine quantitative measures pertaining to predictive accuracy. One of the trained models is selected and then applied over another time series dataset of the individual resource to generate a plurality of individual resource usage predictions. The individual resource usage predictions are used to calculate seasonally-adjusted resource usage demand amounts over a future time period. The resource usage demand amounts are compared to availability of the resource to form a runway that refers to a future time period when the resource is predicted to be demanded to its capacity.

Abstract: A method of generating a question-answer learning model through adversarial learning may include: sampling a latent variable based on constraints in an input passage; generating an answer based on the latent variable; generating a question based on the answer; and machine-learning the question-answer learning model using a dataset of the generated question and answer, wherein the constraints are controlled so that the latent variable is present in a data manifold while increasing a loss of the question-answer learning model.

Abstract: Systems, apparatuses, and methods are provided for identifying a corresponding string stored in memory based on an incomplete input string. A system can analyze and produce phonetic and distance metrics for a plurality of strings stored in memory by comparing the plurality of strings to an incomplete input string. These similarity metrics can be used as the input to a machine learning model, which can quickly and accurately provide a classification. This classification can be used to identify a string stored in memory that corresponds to the incomplete input string.