张量网络是由许多低阶张量组成的网络来表示高阶张量的有效方法,这些低阶张量网络在量子物理和应用数学中都有研究。近年来,TNs在机器学习中得到了越来越多的研究和应用,用于深度神经网络(DNNs)的高维数据分析、模型压缩和高效计算,以及对DNNs表达能力的理论分析。本教程旨在从TN数据表示、参数建模和函数逼近的角度介绍TNs技术应用于机器学习的最新进展。具体来说,我们将介绍TNs的基本模型和算法,无监督学习的典型方法,张量补全,多模态学习以及在DNN, CNN, RNN等中的各种应用。我们还讨论了这一研究领域的新前沿和未来趋势。
Part I. 张量方法数据表示 Tensor Methods for Data Representation
Tensor Train and Tensor Ring Models
Tensor Network Diagram
Latent Convex Tensor Decomposition
Tensor Completion for Missing Values
Tensor Decomposition
Part II. 张量网络深度学习建模Tensor Networks in Deep Learning Modeling
Applications to RNN, LSTM, and Transformer
Multimodal Learning by Tensor Networks
Tensor Networks for Theoretical Analysis of DNNs
Speedup and Compression of CNN
Exponential Machine
Supervised Learning with Quantum Inspired Tensor Networks
Learning Algorithms for Reparametrization by Tensor Networks
Model Compression of NN by Tensor Networks
Part III. 前沿进展与趋势 Frontiers and Future Trends
Discussions
Structure Learning of Tensor Networks
Supervised Learning by Multi-scale TNs, 2D PEPS type TNs, and Tree TNs
Gaussian Mixture Distribution with Multi-dimensional Modes
Generative Modeling by TN
TN Representation for Probabilistic Graphical Model
TN for Function Approximation of Supervised Learning
专知便捷查看
便捷下载,请关注专知公众号(点击上方蓝色专知关注)
后台回复“TML” 可以获取《【ACML2020】张量网络机器学习:最近的进展和前沿,109页ppt》专知下载链接索引