We consider the general $f$-divergence formulation of bidirectional generative modeling, which includes VAE and BiGAN as special cases. We present a new optimization method for this formulation, where the gradient is computed using an adversarially learned discriminator.

With the two-stage fourth-order temporal evolution of the gas distribution function and Weighted Essential Non-Oscillatory (WENO) reconstruction, a high-order finite difference gas-kinetic scheme is proposed.

In random matrix theory, one of the central topics is the limiting behavior of eigenvalues and eigenvectors of random matrices under fixed-rank perturbations. A famous model, raised by Johnstone, is the so-called spiked covariance matrix model.

The polygenic risk score (PRS) derived from the genome-wide association studies (GWASs) predicts the individualized genomic predisposition to complex traits/diseases.

In this paper, the multiscale unified gas-kinetic wave-particle (UGKWP) method has been implemented on three-dimensional unstructured mesh with the capability of large-scale parallel computing.

Cryo-EM detects protein conformations froze in solution, and thus provides a promising way to characterize these conformational changes. In order to detect the conformational change, we developed a new algorithm to obtain multiple conformations and their populations from Cryo-EM datasets.

In this seminar I will talk about a provably stable architecture for Neural Ordinary Differential Equations (ODEs) which achieves non-trivial adversarial robustness under white-box adversarial attacks even when the network is trained naturally.

Federated learning aims to protect data privacy by collaboratively learning a model without sharing private data among users. However, an adversary may still be able to infer the private training data by attacking the released model.