Fundamental research and developments in the field of applied artificial intelligence
The present stage of development of artificial intelligence (AI) is based on advanced
technologies and modern methods and algorithms of machine learning (ML) including deep …
technologies and modern methods and algorithms of machine learning (ML) including deep …
Survey on large scale neural network training
Modern Deep Neural Networks (DNNs) require significant memory to store weight,
activations, and other intermediate tensors during training. Hence, many models do not fit …
activations, and other intermediate tensors during training. Hence, many models do not fit …
Quantization of Large Language Models with an Overdetermined Basis
In this paper, we introduce an algorithm for data quantization based on the principles of
Kashin representation. This approach hinges on decomposing any given vector, matrix, or …
Kashin representation. This approach hinges on decomposing any given vector, matrix, or …
[PDF][PDF] Survey on Efficient Training of Large Neural Networks
DD Shliazhko, I Oseledets, O Beaumont - ijcai.org
Abstract Modern Deep Neural Networks (DNNs) require significant memory to store weight,
activations, and other intermediate tensors during training. Hence, many models don't fit one …
activations, and other intermediate tensors during training. Hence, many models don't fit one …
[引用][C] NOVEL METHODS FOR QUANTIZATION AND ACCELERATION OF NEURAL NETWORKS USING EFFICIENT MATRIX REPRESENTATIONS
D CHERNIUK - 2024