arXiv subjects

Chao Du

Publications and source records attributed to Chao Du.

103 records · Page 6Linked to original sources

arXiv2020

Rethinking Softmax Cross-Entropy Loss for Adversarial Robustness

Previous work shows that adversarially robust generalization requires larger sample complexity, and the same dataset, e.g., CIFAR-10, which enables good standard accuracy may not suffice to train robust models. Since collecting new training data could be costly, we focus on better utilizing the given data by inducing the regions with high sample density in the feature space, which could lead to locally sufficient samples for robust learning. We first formally show that the softmax cross-entropy (SCE) loss and its variants convey inappropriate supervisory signals, which encourage the learned feature points to spread over the space sparsely in training. This inspires us to propose the Max-Mahalanobis center (MMC) loss to explicitly induce dense feature regions in order to benefit robustness. Namely, the MMC loss encourages the model to concentrate on learning ordered and compact representations, which gather around the preset optimal centers for different classes. We empirically demonstrate that applying the MMC loss can significantly improve robustness even under strong adaptive attacks, while keeping state-of-the-art accuracy on clean inputs with little extra computation compared to the SCE loss.

cs.LG↗

arXiv2020

To Relieve Your Headache of Training an MRF, Take AdVIL

We propose a black-box algorithm called {\it Adversarial Variational Inference and Learning} (AdVIL) to perform inference and learning on a general Markov random field (MRF). AdVIL employs two variational distributions to approximately infer the latent variables and estimate the partition function of an MRF, respectively. The two variational distributions provide an estimate of the negative log-likelihood of the MRF as a minimax optimization problem, which is solved by stochastic gradient descent. AdVIL is proven convergent under certain conditions. On one hand, compared with contrastive divergence, AdVIL requires a minimal assumption about the model structure and can deal with a broader family of MRFs. On the other hand, compared with existing black-box methods, AdVIL provides a tighter estimate of the log partition function and achieves much better empirical results.

cs.LG↗

arXiv2019

Improving Adversarial Robustness via Promoting Ensemble Diversity

Though deep neural networks have achieved significant progress on various tasks, often enhanced by model ensemble, existing high-performance models can be vulnerable to adversarial attacks. Many efforts have been devoted to enhancing the robustness of individual networks and then constructing a straightforward ensemble, e.g., by directly averaging the outputs, which ignores the interaction among networks. This paper presents a new method that explores the interaction among individual networks to improve robustness for ensemble models. Technically, we define a new notion of ensemble diversity in the adversarial setting as the diversity among non-maximal predictions of individual members, and present an adaptive diversity promoting (ADP) regularizer to encourage the diversity, which leads to globally better robustness for the ensemble by making adversarial examples difficult to transfer among individual members. Our method is computationally efficient and compatible with the defense methods acting on individual networks. Empirical results on various datasets verify that our method can improve adversarial robustness while maintaining state-of-the-art accuracy on normal examples.

cs.LG↗

arXiv2018

Towards Robust Detection of Adversarial Examples

Although the recent progress is substantial, deep learning methods can be vulnerable to the maliciously generated adversarial examples. In this paper, we present a novel training procedure and a thresholding test strategy, towards robust detection of adversarial examples. In training, we propose to minimize the reverse cross-entropy (RCE), which encourages a deep network to learn latent representations that better distinguish adversarial examples from normal ones. In testing, we propose to use a thresholding strategy as the detector to filter out adversarial examples for reliable predictions. Our method is simple to implement using standard algorithms, with little extra training cost compared to the common cross-entropy minimization. We apply our method to defend various attacking methods on the widely used MNIST and CIFAR-10 datasets, and achieve significant improvements on robust predictions under all the threat models in the adversarial setting.

cs.LG↗

arXiv2018

Collaborative Filtering with User-Item Co-Autoregressive Models

Deep neural networks have shown promise in collaborative filtering (CF). However, existing neural approaches are either user-based or item-based, which cannot leverage all the underlying information explicitly. We propose CF-UIcA, a neural co-autoregressive model for CF tasks, which exploits the structural correlation in the domains of both users and items. The co-autoregression allows extra desired properties to be incorporated for different tasks. Furthermore, we develop an efficient stochastic learning algorithm to handle large scale datasets. We evaluate CF-UIcA on two popular benchmarks: MovieLens 1M and Netflix, and achieve state-of-the-art performance in both rating prediction and top-N recommendation tasks, which demonstrates the effectiveness of CF-UIcA.

cs.LG↗

arXiv2018

Max-Mahalanobis Linear Discriminant Analysis Networks

A deep neural network (DNN) consists of a nonlinear transformation from an input to a feature representation, followed by a common softmax linear classifier. Though many efforts have been devoted to designing a proper architecture for nonlinear transformation, little investigation has been done on the classifier part. In this paper, we show that a properly designed classifier can improve robustness to adversarial attacks and lead to better prediction results. Specifically, we define a Max-Mahalanobis distribution (MMD) and theoretically show that if the input distributes as a MMD, the linear discriminant analysis (LDA) classifier will have the best robustness to adversarial examples. We further propose a novel Max-Mahalanobis linear discriminant analysis (MM-LDA) network, which explicitly maps a complicated data distribution in the input space to a MMD in the latent feature space and then applies LDA to make predictions. Our results demonstrate that the MM-LDA networks are significantly more robust to adversarial attacks, and have better performance in class-biased classification.

cs.LG↗

arXiv2017

Stochastic Modeling and Statistical Inference of Intrinsic Noise in Gene Regulation System via Chemical Master Equation

Intrinsic noise, the stochastic cell-to-cell fluctuations in mRNAs and proteins, has been observed and proved to play important roles in cellular systems. Due to the recent development in single-cell-level measurement technology, the studies on intrinsic noise are becoming increasingly popular among scholars. The chemical master equation (CME) has been used to model the evolutions of complex chemical and biological systems since 1940, and are often served as the standard tool for modeling intrinsic noise in gene regulation system. A CME-based model can capture the discrete, stochastic, and dynamical nature of gene regulation system, and may offer casual and physical explanation of the observed data at single-cell level. Nonetheless, the complexity of CME also pose serious challenge for researchers in proposing practical modeling and inference frameworks. In this article, we will review the existing works on the modelings and inference of intrinsic noise in gene regulation system within the framework of CME model. We will explore the principles in constructing a CME model for studying gene regulation system and discuss the popular approximations of CME. Then we will study the simulation simulation methods as well as the analytical and numerical approaches that can be used to obtain solution to a CME model. Finally we will summary the exiting statistical methods that can be used to infer the unknown parameters or structures in CME model using single-cell-level gene expression data.

stat.ME↗

arXiv2017

Field-Programmable Crossbar Array (FPCA) for Reconfigurable Computing

For decades, advances in electronics were directly driven by the scaling of CMOS transistors according to Moore's law. However, both the CMOS scaling and the classical computer architecture are approaching fundamental and practical limits, and new computing architectures based on emerging devices, such as resistive random-access memory (RRAM) devices, are expected to sustain the exponential growth of computing capability. Here we propose a novel memory-centric, reconfigurable, general purpose computing platform that is capable of handling the explosive amount of data in a fast and energy-efficient manner. The proposed computing architecture is based on a uniform, physical, resistive, memory-centric fabric that can be optimally reconfigured and utilized to perform different computing and data storage tasks in a massively parallel approach. The system can be tailored to achieve maximal energy efficiency based on the data flow by dynamically allocating the basic computing fabric for storage, arithmetic, and analog computing including neuromorphic computing tasks.

cs.ET↗

arXiv2016

Learning Deep Generative Models with Doubly Stochastic MCMC

We present doubly stochastic gradient MCMC, a simple and generic method for (approximate) Bayesian inference of deep generative models (DGMs) in a collapsed continuous parameter space. At each MCMC sampling step, the algorithm randomly draws a mini-batch of data samples to estimate the gradient of log-posterior and further estimates the intractable expectation over hidden variables via a neural adaptive importance sampler, where the proposal distribution is parameterized by a deep neural network and learnt jointly. We demonstrate the effectiveness on learning various DGMs in a wide range of tasks, including density estimation, data generation and missing data imputation. Our method outperforms many state-of-the-art competitors.

cs.LG↗

arXiv2014

Inner Product Similarity Search using Compositional Codes

This paper addresses the nearest neighbor search problem under inner product similarity and introduces a compact code-based approach. The idea is to approximate a vector using the composition of several elements selected from a source dictionary and to represent this vector by a short code composed of the indices of the selected elements. The inner product between a query vector and a database vector is efficiently estimated from the query vector and the short code of the database vector. We show the superior performance of the proposed group $M$-selection algorithm that selects $M$ elements from $M$ source dictionaries for vector approximation in terms of search accuracy and efficiency for compact codes of the same length via theoretical and empirical analysis. Experimental results on large-scale datasets ($1M$ and $1B$ SIFT features, $1M$ linear models and Netflix) demonstrate the superiority of the proposed approach.

cs.CV↗

arXiv2014

3D Vertical Dual-Layer Oxide Memristive Devices for Neuromorphic Computing

Dual-layer resistive switching devices with horizontal W electrodes, vertical Pd electrodes and WOx switching layer formed at the sidewall of the horizontal electrodes have been fabricated and characterized. The devices exhibit well-characterized analog switching characteristics and small mismatch in electrical characteristics for devices formed at the two layers. The three-dimensional (3D) vertical device structure allows higher storage density and larger connectivity for neuromorphic computing applications. We show the vertical devices exhibit potentiation and depression characteristics similar to planar devices, and can be programmed independently with no crosstalk between the layers.

cond-mat.other↗

arXiv2013

Computational Aspects of Optional Pólya Tree

Optional Pólya Tree (OPT) is a flexible non-parametric Bayesian model for density estimation. Despite its merits, the computation for OPT inference is challenging. In this paper we present time complexity analysis for OPT inference and propose two algorithmic improvements. The first improvement, named Limited-Lookahead Optional Pólya Tree (LL-OPT), aims at greatly accelerate the computation for OPT inference. The second improvement modifies the output of OPT or LL-OPT and produces a continuous piecewise linear density estimate. We demonstrate the performance of these two improvements using simulations.

stat.CO↗

arXiv2012

Correlation analysis of enzymatic reaction of a single protein molecule

New advances in nano sciences open the door for scientists to study biological processes on a microscopic molecule-by-molecule basis. Recent single-molecule biophysical experiments on enzyme systems, in particular, reveal that enzyme molecules behave fundamentally differently from what classical model predicts. A stochastic network model was previously proposed to explain the experimental discovery. This paper conducts detailed theoretical and data analyses of the stochastic network model, focusing on the correlation structure of the successive reaction times of a single enzyme molecule. We investigate the correlation of experimental fluorescence intensity and the correlation of enzymatic reaction times, and examine the role of substrate concentration in enzymatic reactions. Our study shows that the stochastic network model is capable of explaining the experimental data in depth.

stat.AP↗