Abstract: An encoder and decoder for translating sequential data into a fixed dimensional vector are created by applying an encoding-trainer input vector set as input to an encoding neural network to generate an encoding-trainer output vector set. One vector of the encoding-trainer output vector set is selected and a decoding-trainer input vector set is generated from it. A decoding neural network is trained by applying the generated decoding-trainer input vector set to the decoding neural network. The encoder and decoder can be used in implementations processing sequential data.

Abstract: Methods, systems, and apparatus for efficiently performing a computation of a convolutional neural network layer. One of the methods includes transforming a X by Y by Z input tensor into a X? by Y? by Z? input tensor, wherein X? is smaller than or equal to X, Y? is smaller than or equal to Y, and Z? is larger than or equal to Z; obtaining one or more modified weight matrices, wherein the modified weight matrices operate on the X? by Y? by Z? input tensor to generate a U? by V? by W? output tensor, and the U? by V? by W? output tensor comprises a transformed U by V by W output tensor; and processing the X? by Y? by Z? input tensor using the modified weight matrices to generate the U? by V? by W? output tensor.

Abstract: Embodiments are disclosed for health assessment and diagnosis implemented in an artificial intelligence (AI) system. In an embodiment, a method comprises: feeding a first set of input features to the AI model; obtaining a first set of raw output predictions from the model; determining a first set of impact scores for the input features fed into the model; training a neural network with the first set of impact scores as input to the network and pre-determined sentences describing the model's behavior as output; feeding a second set of input features to the AI model; obtaining a second set of raw output predictions from the model; determining a second set of impact scores based on the second set of output predictions; feeding the second set of impact scores to the neural network; and generating a sentence describing the AI model's behavior on the second set of input features.

Abstract: A method includes determining a feature importance ranking for each pair of clusters of a plurality of clusters to generate a first plurality of feature importance rankings. The method further includes determining a feature importance ranking between a particular data element and each cluster to generate a second plurality of feature importance rankings. A distance value associated with each pair of clusters of the plurality of clusters is determined to generate a plurality of distance values, and a probability value associated with each data element is determined to generate a plurality of probability values. The method further includes weighting the first plurality of feature importance rankings based on the plurality of distance values to determine a first plurality of weighted feature importance rankings and weighting the second plurality of feature importance rankings based on the plurality of probability values to determine a second plurality of weighted feature importance rankings.

Abstract: Methods are provided for evaluating and/or predicting the outcome of a clinical condition, such as cancer, metastasis, AIDS, autism, Alzheimer's, and/or Parkinson's disorder. The methods can also be used to monitor and track changes in a patient's DNA and/or RNA during and following a clinical treatment regime. The methods may also be used to evaluate protein and/or metabolite levels that correlate with such clinical conditions. The methods are also of use to ascertain the probability outcome for a patient's particular prognosis.

Abstract: Systems, methods, and non-transitory computer readable media configured to acquire data associated with a content item, the data associated with the content item including contextual information. The data associated with the content item can be provided to a model trained by machine learning. A set of hashtags associated with the content item can be determined based on the model.

Abstract: In some examples, a computing device may receive hierarchical data having a hierarchical structure including a plurality of levels. The computing device may determine a plurality of features based at least in part on the hierarchical data, and may select a subset of the features at a first level as candidates for consolidating to a next higher level in the hierarchical structure. The computing device may determine that a predicted loss of information from consolidating the subset of features is less than a threshold, and may revise the hierarchical structure to include a consolidated feature at the next higher level, rather than the subset of features. In some examples, a statistical model may be trained based on the revised hierarchical structure and used at least partially to make a determination, send a notification, and/or control a device.

Abstract: The methods include, for instance: building model connections between models in a knowledgebase, which stores case data as model networks. An exploration probability stored in the knowledgebase indicates a likelihood of new connections based on a case data input to be employed for an analytical model of the case data input, which includes numerous stages and multiple model choices in each stage. Based on the new connections and model networks of the knowledgebase, paths are created and performance of respective paths/connections are evaluated. A predefined number of top performing paths are selected and the models and attributes that do not appear in the top performing paths are eliminated from the knowledgebase. Probabilities of model networks and a future case data input are updated accordingly and a best-fit model for the case data input is presented to a user.

Abstract: A neuromorphic network includes a first node configured to transmit a first optical signal and a second node configured to transmit a second optical signal. A waveguide optically connects the first node to the second node. An integrated optical synapse is located on the waveguide between the first node and the second node, the optical synapse configured to change an optical property based on the first optical signal and the second optical signal such that if a correlation between the first optical signal and the second optical signal is strong, the optical connection between the first node and the second node is increased and if the correlation between the first optical signal and the second optical signal is weak, the optical connection between the first node and the second node is decreased.

Abstract: Systems, apparatus and methods are provided for accelerating a complex neural network by fixed-point data quantization. An Artificial Neural Network (ANN) has branches and comprises convolutional layers CONV 1, CONV 2, . . . CONV n, fully connected layers FC 1, FC 2, . . . , FC m, and concatenation layers CONCAT1, CONCAT2, . . . , CONCAT L. n, m and L are positive integers. The ANN may be optimized by a method comprising: converting output of each of the CONV, FC and CONCAT layers into fixed-point numbers, identifying at least one sub-network from the ANN and for each sub-network, modifying the fixed-point range of each output of the previous-level layers of the CONCAT layer on the basis of the fixed-point range of the CONCAT layer. The sub-network has a CONCAT layer as its output. The CONCAT layer receives at least two outputs of previous-level layers as inputs and concatenates the inputs into one output.

Abstract: Performance testing based on variable length segmentation and clustering of time series data is disclosed. One example is a system including a training module, a performance testing module, and an interface module. The training module generates a trained model to learn characteristics of a system of interest from training time series data by segmenting the training time series data into homogeneous windows of variable length, clustering the segments to identify patterns, and associating each cluster with a cluster score. The performance testing module analyzes system characteristics from testing time series data by receiving the testing time series data, and determining a performance metric for the testing time series data by analyzing the testing time series data based on the trained model. The interface module is communicatively linked to the performance testing module, and provides the performance metric via an interactive graphical user interface.

Abstract: A disclosed machine learning method includes: calculating a first output error between a label and an output in a case where dropout in which values are replaced with 0 is executed for a last layer of a first channel among plural channels in a parallel neural network; calculating a second output error between the label and an output in a case where the dropout is not executed for the last layer of the first channel; and identifying at least one channel from the plural channels based on a difference between the first output error and the second output error to update parameters of the identified channel.

Abstract: A system and method to identify patterns in sets of signals produced during operation of a complex system and combines the identified patterns with records of past conditions to generate operational feedback to one or more machines of the complex system while it operates.

Abstract: Method, system and computer program product for learning classification model. The present invention provides a computer-implemented method for learning a classification model using one or more training data each having a training input and one or more correct labels assigned to the training input, the classification model having a plurality of hidden units and a plurality of output units is provided. The method includes: obtaining a combination of co-occurring labels expected to be appeared together for an input to the classification model; initializing the classification model with preparing a dedicated unit for the combination from among the plurality of the hidden units so as to activate together related output units connected to the dedicated unit among the plurality of the output units, each related output unit corresponding to each co-occurring label in the combination; and training the classification model using the one or more training data.

Abstract: Mechanisms for clarifying an input question are provided. A question is received for generation of an answer. A set of candidate answers is generated based on an analysis of a corpus of information. Each candidate answer has an evidence passage supporting the candidate answer. Based on the set of candidate answers, a determination is made as to whether clarification of the question is required. In response to a determination that clarification of the question is required, a request is sent for user input to clarify the question. User input is received from the computing device in response to the request and at least one candidate answer in the set of candidate answers is selected as an answer for the question based on the user input.

Abstract: A method of diagnosing an orthodontic condition and providing information regarding orthodontic treatment can include analyzing patient data, which can include online activity of an individual, an image of an individual and a combination thereof, receiving at least a portion of patient data onto a server, accessing one or more databases that comprises or has access to at least one of information derived from orthodontic treatments, information derived from a website, and a combination thereof, and instructing at least one computer program to analyze at least a portion of the patient data and identify at least one diagnosis of an orthodontic condition. A system can include one or more computer programs configured to perform a method according to the disclosure.

Abstract: Described is a system for explaining how the human brain represents conceptual knowledge. A semantic model is developed, and a behavioral exam is performed to assess a calibration subject into a cohort and reveal semantic relationships to modify a personalized semantic space developed by the semantic model. Semantic features are extracted from the personalized semantic space. Neural features are extracted from neuroimaging of the human subject. A neuroceptual lattice is created having nodes representing attributes by aligning the semantic features and the neural features. Structures in the neuroceptual lattice are identified to quantify an extent to which the set of neural features represents a target concept. The identified structures are used to interpret conceptual knowledge in the brain of a test subject.

Abstract: A processor-implemented method, system, and/or computer program product generate an intelligent persona agent for use in designing a product. One or more processors input a persona specification into an intelligent persona agent generator. The persona specification describes attributes of a set of model users of a particular type of product, and the intelligent personal agent generator creates an intelligent persona agent that is a software-based version of the set of model users. The intelligent persona agent monitors intermediate design choices taken during a design of a product of the particular type of product by a design team. In response to the intelligent persona agent identifying an intermediate design choice that will lead to a feature that is in conflict with the persona specification of the intelligent persona agent, designers modify the intermediate design choice, which modifies the design of the product in order to create an improved product design.

Abstract: The present disclosure relates to a computer-implemented method for analyzing one or more hyper-parameters for a multi-layer computational structure. The method may include accessing, using at least one processor, input data for recognition. The input data may include at least one of an image, a pattern, a speech input, a natural language input, a video input, and a complex data set. The method may further include processing the input data using one or more layers of the multi-layer computational structure and performing matrix factorization of the one or more layers. The method may also include analyzing one or more hyper-parameters for the one or more layers based upon, at least in part, the matrix factorization of the one or more layers.

Abstract: Embodiments of the present disclosure are directed to methods, computer program products, and computer systems of a robotic apparatus with robotic instructions replicating a food preparation recipe.