1 code implementation • 21 May 2022 • Zhiqi Bu, Jialin Mao, Shiyun Xu
Large convolutional neural networks (CNN) can be difficult to train in the differentially private (DP) regime, since the optimization algorithms require a computationally expensive operation, known as the per-sample gradient clipping.
no code implementations • 25 Feb 2022 • Shiyun Xu, Zhiqi Bu, Pratik Chaudhari, Ian J. Barnett
In order to empower NAM with feature selection and improve the generalization, we propose the sparse neural additive models (SNAM) that employ the group sparsity regularization (e. g. Group LASSO), where each feature is learned by a sub-network whose trainable parameters are clustered as a group.
no code implementations • 30 Nov 2021 • Changjun He, Xiuqiang He, Hua Geng, Huadong Sun, Shiyun Xu
This criterion is proved to be a sufficient stability condition for addressing the effects of the jumps and cosine damping coefficient on the system stability.
1 code implementation • 1 Nov 2020 • Shiyun Xu, Zhiqi Bu
Recent years have witnessed strong empirical performance of over-parameterized neural networks on various tasks and many advances in the theory, e. g. the universal approximation and provable convergence to global minimum.
1 code implementation • 25 Oct 2020 • Zhiqi Bu, Shiyun Xu, Kan Chen
When equipped with efficient optimization algorithms, the over-parameterized neural networks have demonstrated high level of performance even though the loss function is non-convex and non-smooth.