Abstract: The proposed systems and methods are directed to explainability-augmented AI systems. These systems are configured to automatically identify, based on one or more of metadata associated with labels assigned to sample data and responses to AI-system-related questionnaires, one or more reasons that support the decisions made by an AI model in response to user queries. The proposed systems apply natural language processing (NLP) to transform the explainability data (e.g., metadata and questionnaire data) to generate human reader-friendly output that summarizes the reasoning by which the AI system made a specific decision and offer transparency to the AI-decision-making process.

Abstract: A method for compressing a neural network includes: obtaining a neural network including a plurality of parameters to be compressed; dividing the parameters into J blocks; compressing a jth block with Kj compression ratios to generate Kj operation branches; obtaining Kj weighting factors; replacing the jth block with the Kj operation branches weighted by the Kj weighting factors to generate a replacement neural network; performing forward propagation to the replacement neural network, a weighted sum operation being performed on Kj operation results generated by the Kj operation branches with the Kj weighting factors and a result of the operation being used as an output; performing backward propagation to the replacement neural network, updated values of the Kj weighting factors being calculated based on a model loss; and determining an operation branch corresponding to the maximum value of the updated values of the Kj weighting factors as a compressed jth block.

Abstract: A method for compressing a neural network includes: obtaining a neural network including J operation layers; compressing a jth operation layer with Kj compression ratios to generate Kj operation branches; obtaining Kj weighting factors; replacing the jth operation layer with the Kj operation branches weighted by the Kj weighting factors to generate a replacement neural network; performing forward propagation to the replacement neural network, a weighted sum operation being performed on Kj operation results generated by the Kj operation branches with the Kj weighting factors and a result of the weighted sum operation being used as an output of the jth operation layer; performing backward propagation to the replacement neural network, updated values of the Kj weighting factors being calculated based on a model loss; and determining an operation branch corresponding to the maximum value of the updated values of the Kj weighting factors as a compressed jth operation layer.

Abstract: Generating visual workflow representations by receiving data including text instructions, identifying actions in the instructions, generating a mapping of the actions according to a generative model, the mapping including an action sequence, providing the mapping to a user, receiving feedback from the user, altering the generative model according to the feedback, and generating a revised mapping according to the feedback.

Abstract: A neural network operation apparatus and method is provided. The neural network operation apparatus includes a memory configured to store data for a neural network operation, and a processor configured to validate the data based on a determination that the neural network operation should be performed on the data, obtain a real memory address to perform the neural network operation based on a result of the validating and a virtual tensor address of the data, and perform the neural network operation based on the real memory address.

Abstract: Methods and systems for decentralized federated learning are described. Each client participating the training of a local machine learning model identifies one or more neighbor clients in direct communication with itself. Each client transmits to its neighbor clients a weighting coefficient and a set of local model parameters for the local model. Each client also receives from its neighbor clients respective sets of local model parameters and respective weighting coefficients. Each client updates its own set of local model parameters using a weighted aggregation of the received sets of local model parameters, each received set of local model parameters being weighted with the respective received weighting coefficient. Each client trains its local machine learning model using a machine learning algorithm and its own local dataset.