Xinyun Chen

Learn More
An intriguing property of deep neural networks is the existence of adversarial examples , which can transfer among different architectures. These transferable ad-versarial examples may severely hinder deep neural network-based applications. Previous works mostly study the transferability using small scale datasets. In this work, we are the first to conduct(More)
This paper develops the first class of algorithms that enable un-biased estimation of steady-state expectations for multidimensional reflected Brownian motion. In order to explain our ideas, we first consider the case of compound Poisson (possibly Markov modulated) input. In this case, we analyze the complexity of our procedure as the dimension of the(More)
Automatic translation from natural language descriptions into programs is a long-standing challenging problem. In this work, we consider a simple yet important sub-problem: translation from textual descriptions to If-Then programs. We devise a novel neural network architecture for this task which we train end-to-end. Specifically, we introduce Latent(More)
When a company undertakes e-commerce transactions for the first time, most major web sites set the initial credit score of the company as zero, which making buyers and sellers can't judge the partners' credibility. In recent years, although commercial banks and some specialized credit rating agencies have established more comprehensive and scientific(More)
Traditional classification algorithms assume that training and test data come from similar distributions. This assumption is violated in adversarial settings, where malicious actors modify instances to evade detection. A number of custom methods have been developed for both adversarial evasion attacks and robust learning. We propose the first systematic and(More)
  • Xinyun Chen
  • 2014
In this paper, we develop a new simulation algorithm that generates unbiased gradient estimators for the steady-state workload of a stochastic fluid network, with respect to the throughput rate of each server. Our algorithm is based on the perfect sampling algorithm developed in Blanchet and Chen (2014), and the infinitesimal perturbation analysis (IPA)(More)
Consider a multidimensional diffusion process X = {X(t) : t ∈ [0, 1]}. Let ε > 0 be a deterministic, user defined, tolerance error parameter. Under standard regularity conditions on the drift and diffusion coefficients of X, we construct a probability space, supporting both X and an explicit, piecewise constant, fully simulatable process X ε such that sup(More)