Abstract: Circuit-Under-Pad (CUP) device topologies for high current lateral GaN power transistors comprise source, drain and gate finger electrodes on active regions of a plurality of sections of a multi-section transistor, and a contact structure comprising source and drain contact areas, e.g. drain and source pads extending over active regions of each section, interconnected by conductive micro-vias to respective underlying drain and source finger electrodes. Alternatively, source contact areas comprise parts of a source bus which runs over inactive regions. For reduced gate loop inductance, the source bus may be routed over or under the to gate bus. The pad structure and the micro-via interconnections are configured to reduce current density in self-supported widths of the drain finger electrodes. Example CUP device structures provide for higher current carrying capability and reduced drain-source resistance.

Abstract: Decentralized bilevel optimization techniques for personalized learning over a heterogenous network are provided. In one aspect, a decentralized learning system includes: a distributed machine learning network with multiple nodes, and datasets associated with the nodes; and a bilevel learning structure at each of the nodes for optimizing one or more features from each of the datasets using a decentralized bilevel optimization solver, while maintaining distinct features from each of the datasets. A method for decentralized learning is also provided.

Abstract: Computer hardware and/or software that performs the following operations: (i) updating a machine learning model by synchronously applying, to the machine learning model, a first set of training results received from a set of trainers having respective training datasets; (ii) receiving, from one or more trainers of the set of trainers, a first set of metrics pertaining to at least some of the training results of the first set of training results; and (iii) based, at least in part, on the first set of metrics, determining to subsequently update the machine learning model via asynchronous application of subsequent training results received from respective trainers of the set of trainers.

Abstract: A method, computer system, and a computer program product are provided for a context-aware relevancy modelling in conversational systems. A user query is received. A latent static content d is selected from a corpus of content D. A latent set of context C from a set of external context Cu is also selected. A result is generated using a scoring function and using the latent static content d from a corpus D and the latent set of context C from the set of external contexts CU so as to provide a most relevant context-base search response to said user query q. The result provides a most relevant context-base search response to said user query q. A response is then generated based on said result using said scoring function result to said user query q.

Abstract: Disclosed in this application are a method for embedding a watermark in video data and apparatus, a method for extracting a watermark in video data and apparatus, a device, and a storage medium. The method for embedding the watermark includes: acquiring a target image frame in video data; performing time-frequency transformation on the target image frame to obtain target frequency domain data, the target frequency domain data comprising a matrix formed by frequency domain coefficients; changing the frequency domain coefficients in the target frequency domain data according to watermark data to obtain watermarked frequency domain data; performing inverse time-frequency transformation on the watermarked frequency domain data to obtain a watermarked image frame; and synthesizing watermarked video data according to the watermarked image frame.

Abstract: Systems, computer-implemented methods, and/or computer program products to facilitate updating, such as averaging and/or training, of one or more statistical sets are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can include a computing component that updates a first statistical set with an additional statistical set from an additional system. The additional statistical set can have been generated from a parent statistical set that is based on underlying data. To update the first statistical set, the additional statistical set can be obtained by the system without obtaining the parent statistical set and without obtaining the underlying data. According to an embodiment, the first statistical set can be a model parameter set generated from a first parent statistical set that is an analytical model.

Abstract: A backbone model parameter and a classification head parameter are randomly initialized. A gradient descent is applied to a lower-level unsupervised loss with respect to the initialized backbone model parameter and the initialized backbone model parameter is updated. A gradient descent is applied to a higher-level supervised loss and the initialized classification head parameter is updated. Deployment of the updated backbone model parameter and the updated classification head parameter are facilitated.

Abstract: One or more systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to length perturbation techniques for improving generalization of DNN acoustic models. A computer-implemented system can comprise a memory that can store computer executable components. The computer-implemented system can further comprise a processor that can execute the computer executable components stored in the memory, wherein the computer executable components can comprise a frame skipping component that can remove one or more frames from an acoustic utterance via frame skipping. The computer executable components can further comprise a frame insertion component that can insert one or more replacement frames into the acoustic utterance via frame insertion to replace the one or more frames with the one or more replacement frames to enable length perturbation of the acoustic utterance.

Abstract: Hessian matrix-free sample-based techniques for model explanations that are faithful to the model are provided. In one aspect, a method for explaining a machine learning model {circumflex over (?)} (e.g., for natural language processing) is provided. The method includes: training the machine learning model {circumflex over (?)} with training data D; obtaining a decision of the machine learning model {circumflex over (?)}; and explaining the decision of the machine learning model {circumflex over (?)} using training examples from the training data D.

Abstract: A machine learning model can be trained to predict one or more financial indicators using earnings call transcripts augmented with counterfactual information. Using faithful gradient-based method, prediction results with respect to a particular counterfactual information can be explained. Based on the explanation, the counterfactual information determined to have most impact on prediction results can be selected for updating the machine learning model.

Abstract: Disclosed is a sediment classification method and system based on bottom echo information of a deep-sea multibeam water column, falling within the technical field of seabed sediment classification. Multidimensional features of seabed multibeam time series are extracted by using multibeam water column data, and multidimensional features of seabed spots are constructed by fusing an angle response curve; double-waveform spectrum comparison of original sample point waveform with seabed measured point waveform is performed, and two groups of corresponding waveforms are matched by using Cosine Similarity to expand original sample points; and the reliability of classification results is improved. In the present disclosure, the transformation from hard to soft classification is realized, and the seabed habitat inversion of complex mixed sediment in a deep sea is further realized.

Abstract: A method, computer system, and a computer program product are provided for a context-aware relevancy modelling in conversational systems. A user query is received. A latent static content d is selected from a corpus of content D. A latent set of context C from a set of external context Cu is also selected. A result is generated using a scoring function and using the latent static content d from a corpus D and the latent set of context C from the set of external contexts CU so as to provide a most relevant context-base search response to said user query q. The result provides a most relevant context-base search response to said user query q. A response is then generated based on said result using said scoring function result to said user query q.

Abstract: A computer-implemented method of decentralized multi-agent learning for use in a system having a plurality of intelligent agents each having a personal portion and a shared portion, is provided. The method includes iteratively, until each of a personal goal and a network goal are optimized: determining a feedback associated with an action relative to a personal goal and a degree of similarity relative to a shared goal; adjusting a policy based on the feedback to gain a superior feedback from a next action; broadcasting the shared policy; receiving the at least one of the one or more other intelligent agents' shared policy; generating a combined policy by combining the personal policy and the at least one of the one or more other intelligent agents' shared policy; estimating, using the combined policy, a network value function; and conducting the next action in accordance with the combined policy.

Abstract: Mechanisms are provided for performing machine learning training of a computer model. A perturbation generator generates a modified training data comprising perturbations injected into original training data, where the perturbations cause a data corruption of the original training data. The modified training data is input into a prediction network of the computer model and processing the modified training data through the prediction network to generate a prediction output. Machine learning training is executed of the prediction network based on the prediction output and the original training data to generate a trained prediction network of a trained computer model. The trained computer model is deployed to an artificial intelligence computing system for performance of an inference operation.

Abstract: One or more systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to length perturbation techniques for improving generalization of DNN acoustic models. A computer-implemented system can comprise a memory that can store computer executable components. The computer-implemented system can further comprise a processor that can execute the computer executable components stored in the memory, wherein the computer executable components can comprise a frame skipping component that can remove one or more frames from an acoustic utterance via frame skipping. The computer executable components can further comprise a frame insertion component that can insert one or more replacement frames into the acoustic utterance via frame insertion to replace the one or more frames with the one or more replacement frames to enable length perturbation of the acoustic utterance.

Abstract: One or more systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to n-best based label smoothing techniques for improving generalization of DNN acoustic models. A computer-implemented system can comprise a memory that can store computer executable components. The computer-implemented system can further comprise a processor that can execute the computer executable components stored in the memory, wherein the computer executable components can comprise a generation component that can generate one or more n-best hypotheses of a ground truth label sequence, using one or more acoustic models, wherein the one or more n-best hypotheses of the ground truth label sequence can represent one or more competing labels that can be used to smooth out the ground truth label sequence.

Abstract: Embodiments of a method are disclosed. The method includes performing distributed deep learning training on multiple batches of training data using corresponding learners. Additionally, the method includes determining training times wherein the learners perform the distributed deep learning training on the batches of training data. The method also includes modifying a processing aspect of the straggler to reduce a future training time of the straggler for performing the distributed deep learning training on a new batch of training data in response to identifying a straggler of the learners by a centralized control.

Abstract: A method for power saving and encryption during analysis of media captured by an information capture device using a partitioned neural network includes replicating, by an information capture device, an artificial neural network (ANN) from a computer server to the information capture device. The ANN on the computer server and a replicated ANN, both, include M layers. The method further includes, in response to captured data being input to be processed, partially processing, by the information capture device, the captured data by executing a first k layers using the replicated ANN, wherein only the k layers are selected to execute on the information capture device. The method further includes transmitting, by the information capture device, an output of the k-th layer to the computer server, which partially processes the captured data by executing the remainder of the M layers using the ANN and the output of the k-th layer.

Abstract: Decentralized bilevel optimization techniques for personalized learning over a heterogenous network are provided. In one aspect, a decentralized learning system includes: a distributed machine learning network with multiple nodes, and datasets associated with the nodes; and a bilevel learning structure at each of the nodes for optimizing one or more features from each of the datasets using a decentralized bilevel optimization solver, while maintaining distinct features from each of the datasets. A method for decentralized learning is also provided.

Abstract: From metadata of a corpus of natural language text documents, a relativity matrix is constructed, a row-column intersection in the relativity matrix corresponding to a relationship between two instances of a type of metadata. An encoder model is trained, generating a trained encoder model, to compute an embedding corresponding to a token of a natural language text document within the corpus and the relativity matrix, the encoder model comprising a first encoder layer, the first encoder layer comprising a token embedding portion, a relativity embedding portion, a token self-attention portion, a metadata self-attention portion, and a fusion portion, the training comprising adjusting a set of parameters of the encoder model.