Abstract: A reinforcement learning ranker can take into account previously-recommended media content items to produce a ranked list of media content items to recommend next. The ranker finds a policy that gives the probability of sampling a media content item given a state. The policy is learned such that it maximizes a reward. A reward function associated with the media content item can be defined with respect to whether the user finds the media content item relevant (likelihood that the user will like the media content item) and a diversity score of the media content item.

Abstract: A system and/or a method for a cognitive platform for autonomous data orchestration, comprising edge computing devices, data platforms and cognitive computing engine. The cognitive computing engine is configured to receive digital events either generated by external computing devices or within the cognitive computing engine. Context and state information is accessed from a central intelligence store to determine the intent of the received digital events. A target entity is identified based on at least one of the determined intent. A data package is composed for the identified at least one target entity, transformed to a natural language text and communicated to the identified at least one target entity, inciting the target entity to take action on the package delivered. The central intelligence store is updated new context and intents for subsequent digital events making it an autonomous learning platform.

Abstract: A method for monitoring operator compatibility within a deep learning framework, a computing system, and a non-transitory computer-readable storage medium are provided. The present disclosure relates to the field of deep learning. The method includes: generating first description information associated with at least one original operator and second description information associated with at least one modified operator, determining differences between the first description information associated with the at least one original operator and the second description information associated with the at least one modified operator, determining whether the differences satisfy a preset rule; and prompting information about incompatibility in response to determining that at least one of the differences does not satisfy the preset rule. The first and second description information are associated with the operator compatibility.

Abstract: A robot system includes a selection module configured to select a stored demonstration for a robot from a database of stored demonstrations for different tasks of the robot; an encoder module of an attention model, the encoder module configured to determine a similarity value reflecting a similarity between: a user input demonstration for the robot; and the stored demonstration for the robot; and an indicator module configured to indicate whether the stored demonstration is the same as the user input demonstration and belongs to the same task based on the similarity value.

Abstract: A distillation system extracts knowledge from a large pre-trained sequence-to-sequence neural transformer model into a smaller bi-encoder. The pre-trained sequence-to-sequence neural transformer model is trained to translate data from a first domain into a second domain on a large corpus. A teacher model is generated from the pre-trained model by fine-tuning the pre-trained neural transformer model on a smaller translation task with true translation pairs. The fine-tuned model is then used to generate augmented data values which are used with the true translation pairs to train the bi-encoder. The bi-encoder is used for perform cross-domain searches.

Abstract: A computer-implemented method comprising, automatically: analyzing a machine learning dataset which comprises multiple datapoints, to deduce constraints on features of the datapoints; generating a first set of CSP (Constraint Satisfaction Problem) rules expressing the constraints; based on a machine learning model which was trained on the dataset, generating a second set of CSP rules that define one or more perturbation candidates among the features of one of the datapoints; formulating a CSP based on the first and second sets of CSP rules; solving the formulated CSP using a solver; and using the solution of the CSP as a counterfactual explanation of a prediction made by the machine learning model with respect to the one datapoint.

Abstract: A data processing system for generating predictive maintenance models is disclosed, including one or more processors, a memory including one or more digital storage devices, and a plurality of instructions stored in the memory. The instructions are executable by the one or more processors to receive a historical dataset relating to each system of a plurality of systems, the historical dataset including maintenance data and operational data. The instructions are further executable to receive a rule set for processing a first attribute of the operation data and calculate a custom data feature from the historical dataset according to the received rule set. The instructions are further executable to generate a predictive maintenance model, using the custom data feature according to a machine learning method.

Abstract: A system, computer program product, and method are provided for probing model signal awareness. An iterative process is employed to systematically isolate one or more relevant tokens of an input sequence to generate a reduced input sequence. The reduced input sequence is validated and presented to a trained artificial intelligence (AI) model and prediction output is generated. The reduction process is continued while the prediction output stays the same as that of the input sequence, and until a minimal sub-sequence is identified. A signal existence in the minimal sub-sequence is verified and signal awareness of the trained AI model is evaluated. The evaluation includes measuring the verified signal existence against an original signal from the input sentence.

Abstract: A computing system can obtain a plurality of datasets that respectively correspond to a plurality of views of a graph network comprising a plurality of nodes, wherein the plurality of datasets comprise one or more partial view datasets for one or more partial views that comprise data for only a respective subset of the plurality of nodes of the graph network. The computing system can determine, based at least in part on an objective function, a plurality of respective embeddings associated respectively with the plurality of nodes, such as a common embedding set that contains respective embeddings for a common subset of the plurality of nodes that are common among all of the plurality of views, and one or more independent embedding sets that contain respective embeddings for one or more respective subsets of the plurality of nodes described only by a subset of the plurality of views.

Abstract: A method for adapting to a new domain an AI model pre-trained for a current domain. Having at least one main block of the AI model for modeling a target variable and having at least one covariates block of the AI model for modeling covariates effect on the target variable in the current domain, the method comprises: replacing the covariates block with a new covariates block adapted to the new domain, the new covariates block modifying one or more first layers compared to the covariate block, the target variable in the new domain being affected differently by at least one of the one or more covariates; training the new covariates block of the AI model using a new-domain-specific dataset from the new domain; and fine-tuning the at least one main block of the AI model using the new-domain-specific dataset from the new domain.

Abstract: Methods, computer program products, and/or systems are provided that perform the following operations: obtaining asset data; determining an asset class associated with the asset data; initializing a new neural network model, wherein the new neural network model is initialized based on a pretrained model associated with the asset class; training the new neural network model based, at least in part, on the asset data to obtain a trained remaining useful life model; and deploying the trained remaining useful life model to generate prediction data for one or more assets as output of the trained remaining useful life model.

Abstract: A system, computer program product, and method are presented for enriching existing legacy expert systems through refinement of existing rules therein. The method includes identifying a legacy expert system to be enriched, relevant training data, existing rules embedded within the legacy expert system, and, for each existing rule, one or more antecedent factors. The method also includes determining the existing rules do not meet a threshold value for established quality requirements, thereby identifying one or more low-quality rules. The method further includes identifying frequent sets of antecedent factors associated with each low-quality rule, where each frequent set of antecedent factors is established as a frequent set through at least meeting a threshold frequency of occurrence within the training data. The method also includes comparing the antecedent factors of each existing rule with the frequent sets of antecedent factors, and enriching the legacy expert system through refining the existing rules.

Abstract: Techniques for providing an overlap data buffer to store portions of tiles between passes of chained layers of a neural network are described. One accelerator circuit includes one or more processing units to execute instructions corresponding to the chained layers in multiple passes. In a first pass, the processing unit(s) receives a first input tile of an input feature map from a primary buffer and performs a first operation on the first input tile to obtain a first output tile. The processing unit stores the first output tile in the primary buffer and identifies a portion of the first output tile as corresponding to overlap data between tiles of the input feature map. The processing unit stores the portion in a secondary buffer. In a second pass, the processing unit retrieves the portion to avoid fetching the portion that overlaps and computing the overlap data again.

Abstract: Methods and systems are provided for compressing a deep neural network (NN) model using knowledge distillation. The method includes training a student NN model to minimize a first loss between student model output values generated by the student NN model for a set of original input values and teacher model output values generated by a teacher NN model for the set of original input values, generating, for at least some of the original input values, a respective perturbed value that maximizes a second loss between an output value generated by the student NN model and an output value generated by the teacher NN model, adding the perturbed values to the set of original input values to provide a set of augmented input values and retraining the student NN model using the set of augmented input values.

Abstract: Computer-implemented method for determining continuous information on an expected trajectory of an object, the method comprising at least the following steps carried out by computer hardware components: determining data related to an expected trajectory of an object; and determining at least one parameter value for a continuous function on the basis of the data, wherein the continuous function and the at least one parameter value represent continuous information on the expected trajectory of the object.

Abstract: A mobile device detects a crash event using one or more sensors of a mobile device. The mobile device records a first set of data from the one or more sensors of the mobile device. The mobile device generates a first feature vector including the first set of data and vehicle data that includes an identifier of a vehicle. The mobile device generates a second feature vector using the first set of data and additional data types. The mobile device predicts a confidence of a total loss event by generating a first confidence value from a first machine-learning model using the first feature vector and a second confidence value from a second machine-learning model using the second feature vector.

Abstract: A neuromorphic computing system configured to be trained using unsupervised learning through distributed computing circuits. The neuromorphic computing system comprises an artificial neural network implemented as a grid of locally connected cells wherein each cell comprises hardware components for neural computing and storage, and is connected to its direct closest neighbors. The neuromorphic computing system comprises a clock system providing periodic active clock edges allowing in each cell to simultaneously and synchronously compute the neuron's Euclidean distance to the input, then compute the Best Matching Unit and the Manhattan distance to it in multiple clock cycles based on a time to Manhattan distance transformation, and finally update the neuron's weights.

Abstract: Herein are machine learning techniques that adjust reconstruction loss of a reconstructive model such as an autoencoder based on importances of values of features. In an embodiment and before, during, or after training, the reconstructive model that more or less accurately reconstructs its input, a computer measures, for each distinct value of each feature, a respective importance that is not based on the reconstructive model. For example, importance may be based solely on a training corpus. For each feature during or after training, a respective original loss from the reconstructive model measures a difference between a value of the feature in an input and a reconstructed value of the feature generated by the reconstructive model. For each feature, the respective importance of the input value of the feature is applied to the respective original loss to generate a respective weighted loss. The weighted losses of the features of the input are collectively detected as anomalous or non-anomalous.

Abstract: There is provided a computer-implemented method of training a neural network, the neural network comprising a plurality of interconnected nodes arranged in a plurality of layers and having a plurality of weights associated therewith. The method comprises obtaining, at a computing device, a training dataset from a database, performing, at the computing device, a forward propagation on the training dataset to produce an output of the neural network, and performing, at the computing device, a backward propagation on the output of the neural network to update the plurality of weights, the backward propagation performed using an estimation of a polarity and a neutrality of gradients of the neural network.

Abstract: A visualization recommendation system generates recommendation scores for multiple visualizations that combine data attributes of a dataset with visualization configurations. The visualization recommendation system maps meta-features of the dataset to a meta-feature space and configuration attributes of the visualization configurations to a configuration space. The visualization recommendation system generates meta-feature vectors that describe the mapped meta-features, and generates configuration attribute sets that describe the attributes of the visualization configurations. The visualization recommendation system applies multiple scoring models to the meta-feature vectors and configuration attribute sets, including a wide scoring model and a deep scoring model. In some cases, the visualization recommendation system trains the multiple scoring models using the meta-feature vectors and configuration attribute sets.