arXiv subjects

Jian Peng

Publications and source records attributed to Jian Peng.

At least 109 records · Page 6Linked to original sources

arXiv2018

Action-depedent Control Variates for Policy Optimization via Stein's Identity

Policy gradient methods have achieved remarkable successes in solving challenging reinforcement learning problems. However, it still often suffers from the large variance issue on policy gradient estimation, which leads to poor sample efficiency during training. In this work, we propose a control variate method to effectively reduce variance for policy gradient methods. Motivated by the Stein's identity, our method extends the previous control variate methods used in REINFORCE and advantage actor-critic by introducing more general action-dependent baseline functions. Empirical studies show that our method significantly improves the sample efficiency of the state-of-the-art policy gradient approaches.

stat.ML↗

arXiv2017

Empower Sequence Labeling with Task-Aware Neural Language Model

Linguistic sequence labeling is a general modeling approach that encompasses a variety of problems, such as part-of-speech tagging and named entity recognition. Recent advances in neural networks (NNs) make it possible to build reliable models without handcrafted features. However, in many cases, it is hard to obtain sufficient annotations to train these models. In this study, we develop a novel neural framework to extract abundant knowledge hidden in raw texts to empower the sequence labeling task. Besides word-level knowledge contained in pre-trained word embeddings, character-aware neural language models are incorporated to extract character-level knowledge. Transfer learning techniques are further adopted to mediate different components and guide the language model towards the key knowledge. Comparing to previous methods, these task-specific knowledge allows us to adopt a more concise model and conduct more efficient training. Different from most transfer learning methods, the proposed framework does not rely on any additional supervision. It extracts knowledge from self-contained order information of training sequences. Extensive experiments on benchmark datasets demonstrate the effectiveness of leveraging character-level knowledge and the efficiency of co-training. For example, on the CoNLL03 NER task, model training completes in about 6 hours on a single GPU, reaching F1 score of 91.71$\pm$0.10 without using any extra annotation.

cs.CL↗

arXiv2017

Efficient Localized Inference for Large Graphical Models

We propose a new localized inference algorithm for answering marginalization queries in large graphical models with the correlation decay property. Given a query variable and a large graphical model, we define a much smaller model in a local region around the query variable in the target model so that the marginal distribution of the query variable can be accurately approximated. We introduce two approximation error bounds based on the Dobrushin's comparison theorem and apply our bounds to derive a greedy expansion algorithm that efficiently guides the selection of neighbor nodes for localized inference. We verify our theoretical bounds on various datasets and demonstrate that our localized inference algorithm can provide fast and accurate approximation for large graphical models.

stat.ML↗

arXiv2017

Energy-efficient Amortized Inference with Cascaded Deep Classifiers

Deep neural networks have been remarkable successful in various AI tasks but often cast high computation and energy cost for energy-constrained applications such as mobile sensing. We address this problem by proposing a novel framework that optimizes the prediction accuracy and energy cost simultaneously, thus enabling effective cost-accuracy trade-off at test time. In our framework, each data instance is pushed into a cascade of deep neural networks with increasing sizes, and a selection module is used to sequentially determine when a sufficiently accurate classifier can be used for this data instance. The cascade of neural networks and the selection module are jointly trained in an end-to-end fashion by the REINFORCE algorithm to optimize a trade-off between the computational cost and the predictive accuracy. Our method is able to simultaneously improve the accuracy and efficiency by learning to assign easy instances to fast yet sufficiently accurate classifiers to save computation and energy cost, while assigning harder instances to deeper and more powerful classifiers to ensure satisfiable accuracy. With extensive experiments on several image classification datasets using cascaded ResNet classifiers, we demonstrate that our method outperforms the standard well-trained ResNets in accuracy but only requires less than 20% and 50% FLOPs cost on the CIFAR-10/100 datasets and 66% on the ImageNet dataset, respectively.

cs.LG↗

arXiv2017

On the Selective and Invariant Representation of DCNN for High-Resolution Remote Sensing Image Recognition

Human vision possesses strong invariance in image recognition. The cognitive capability of deep convolutional neural network (DCNN) is close to the human visual level because of hierarchical coding directly from raw image. Owing to its superiority in feature representation, DCNN has exhibited remarkable performance in scene recognition of high-resolution remote sensing (HRRS) images and classification of hyper-spectral remote sensing images. In-depth investigation is still essential for understanding why DCNN can accurately identify diverse ground objects via its effective feature representation. Thus, we train the deep neural network called AlexNet on our large scale remote sensing image recognition benchmark. At the neuron level in each convolution layer, we analyze the general properties of DCNN in HRRS image recognition by use of a framework of visual stimulation-characteristic response combined with feature coding-classification decoding. Specifically, we use histogram statistics, representational dissimilarity matrix, and class activation mapping to observe the selective and invariance representations of DCNN in HRRS image recognition. We argue that selective and invariance representations play important roles in remote sensing images tasks, such as classification, detection, and segment. Also selective and invariance representations are significant to design new DCNN liked models for analyzing and understanding remote sensing images.

cs.CV↗

arXiv2017

What do We Learn by Semantic Scene Understanding for Remote Sensing imagery in CNN framework?

Recently, deep convolutional neural network (DCNN) achieved increasingly remarkable success and rapidly developed in the field of natural image recognition. Compared with the natural image, the scale of remote sensing image is larger and the scene and the object it represents are more macroscopic. This study inquires whether remote sensing scene and natural scene recognitions differ and raises the following questions: What are the key factors in remote sensing scene recognition? Is the DCNN recognition mechanism centered on object recognition still applicable to the scenarios of remote sensing scene understanding? We performed several experiments to explore the influence of the DCNN structure and the scale of remote sensing scene understanding from the perspective of scene complexity. Our experiment shows that understanding a complex scene depends on an in-depth network and multiple-scale perception. Using a visualization method, we qualitatively and quantitatively analyze the recognition mechanism in a complex remote sensing scene and demonstrate the importance of multi-objective joint semantic support.

cs.CV↗

arXiv2017

Stein Variational Policy Gradient

Policy gradient methods have been successfully applied to many complex reinforcement learning problems. However, policy gradient methods suffer from high variance, slow convergence, and inefficient exploration. In this work, we introduce a maximum entropy policy optimization framework which explicitly encourages parameter exploration, and show that this framework can be reduced to a Bayesian inference problem. We then propose a novel Stein variational policy gradient method (SVPG) which combines existing policy gradient methods and a repulsive functional to generate a set of diverse but well-behaved policies. SVPG is robust to initialization and can easily be implemented in a parallel manner. On continuous control problems, we find that implementing SVPG on top of REINFORCE and advantage actor-critic algorithms improves both average return and data efficiency.

cs.LG↗

arXiv2017

On the Interpretability of Conditional Probability Estimates in the Agnostic Setting

We study the interpretability of conditional probability estimates for binary classification under the agnostic setting or scenario. Under the agnostic setting, conditional probability estimates do not necessarily reflect the true conditional probabilities. Instead, they have a certain calibration property: among all data points that the classifier has predicted P(Y = 1|X) = p, p portion of them actually have label Y = 1. For cost-sensitive decision problems, this calibration property provides adequate support for us to use Bayes Decision Theory. In this paper, we define a novel measure for the calibration property together with its empirical counterpart, and prove an uniform convergence result between them. This new measure enables us to formally justify the calibration property of conditional probability estimations, and provides new insights on the problem of estimating and calibrating conditional probabilities.

cs.LG↗

arXiv2016

Learning to Play in a Day: Faster Deep Reinforcement Learning by Optimality Tightening

We propose a novel training algorithm for reinforcement learning which combines the strength of deep Q-learning with a constrained optimization approach to tighten optimality and encourage faster reward propagation. Our novel technique makes deep reinforcement learning more practical by drastically reducing the training time. We evaluate the performance of our approach on the 49 games of the challenging Arcade Learning Environment, and report significant improvements in both training time and accuracy.

cs.LG↗

arXiv2016

Meta-Path Guided Embedding for Similarity Search in Large-Scale Heterogeneous Information Networks

Most real-world data can be modeled as heterogeneous information networks (HINs) consisting of vertices of multiple types and their relationships. Search for similar vertices of the same type in large HINs, such as bibliographic networks and business-review networks, is a fundamental problem with broad applications. Although similarity search in HINs has been studied previously, most existing approaches neither explore rich semantic information embedded in the network structures nor take user's preference as a guidance. In this paper, we re-examine similarity search in HINs and propose a novel embedding-based framework. It models vertices as low-dimensional vectors to explore network structure-embedded similarity. To accommodate user preferences at defining similarity semantics, our proposed framework, ESim, accepts user-defined meta-paths as guidance to learn vertex vectors in a user-preferred embedding space. Moreover, an efficient and parallel sampling-based optimization algorithm has been developed to learn embeddings in large-scale HINs. Extensive experiments on real-world large-scale HINs demonstrate a significant improvement on the effectiveness of ESim over several state-of-the-art algorithms as well as its scalability.

cs.SI↗

arXiv2016

DPPred: An Effective Prediction Framework with Concise Discriminative Patterns

In the literature, two series of models have been proposed to address prediction problems including classification and regression. Simple models, such as generalized linear models, have ordinary performance but strong interpretability on a set of simple features. The other series, including tree-based models, organize numerical, categorical and high dimensional features into a comprehensive structure with rich interpretable information in the data. In this paper, we propose a novel Discriminative Pattern-based Prediction framework (DPPred) to accomplish the prediction tasks by taking their advantages of both effectiveness and interpretability. Specifically, DPPred adopts the concise discriminative patterns that are on the prefix paths from the root to leaf nodes in the tree-based models. DPPred selects a limited number of the useful discriminative patterns by searching for the most effective pattern combination to fit generalized linear models. Extensive experiments show that in many scenarios, DPPred provides competitive accuracy with the state-of-the-art as well as the valuable interpretability for developers and experts. In particular, taking a clinical application dataset as a case study, our DPPred outperforms the baselines by using only 40 concise discriminative patterns out of a potentially exponentially large set of patterns.

cs.LG↗

arXiv2016

Complexes Detection in Biological Networks via Diversified Dense Subgraphs Mining

Protein-protein interaction (PPI) networks, providing a comprehensive landscape of protein interacting patterns, enable us to explore biological processes and cellular components at multiple resolutions. For a biological process, a number of proteins need to work together to perform the job. Proteins densely interact with each other, forming large molecular machines or cellular building blocks. Identification of such densely interconnected clusters or protein complexes from PPI networks enables us to obtain a better understanding of the hierarchy and organization of biological processes and cellular components. Most existing methods apply efficient graph clustering algorithms on PPI networks, often failing to detect possible densely connected subgraphs and overlapped subgraphs. Besides clustering-based methods, dense subgraph enumeration methods have also been used, which aim to find all densely connected protein sets. However, such methods are not practically tractable even on a small yeast PPI network, due to high computational complexity. In this paper, we introduce a novel approximate algorithm to efficiently enumerate putative protein complexes from biological networks. The key insight of our algorithm is that we do not need to enumerate all dense subgraphs. Instead we only need to find a small subset of subgraphs that cover as many proteins as possible. The problem is formulated as finding a diverse set of dense subgraphs, where we develop highly effective pruning techniques to guarantee efficiency. To handle large networks, we take a divide-and-conquer approach to speed up the algorithm in a distributed manner. By comparing with existing clustering and dense subgraph-based algorithms on several human and yeast PPI networks, we demonstrate that our method can detect more putative protein complexes and achieves better prediction accuracy.

q-bio.MN↗

arXiv2016

Low-density locality-sensitive hashing boosts metagenomic binning

Metagenomic binning is an essential task in analyzing metagenomic sequence datasets. To analyze structure or function of microbial communities from environmental samples, metagenomic sequence fragments are assigned to their taxonomic origins. Although sequence alignment algorithms can readily be used and usually provide high-resolution alignments and accurate binning results, the computational cost of such alignment-based methods becomes prohibitive as metagenomic datasets continue to grow. Alternative compositional-based methods, which exploit sequence composition by profiling local short k-mers in fragments, are often faster but less accurate than alignment-based methods. Inspired by the success of linear error correcting codes in noisy channel communication, we introduce Opal, a fast and accurate novel compositional-based binning method. It incorporates ideas from Gallager's low-density parity-check code to design a family of compact and discriminative locality-sensitive hashing functions that encode long-range compositional dependencies in long fragments. By incorporating the Gallager LSH functions as features in a simple linear SVM, Opal provides fast, accurate and robust binning for datasets consisting of a large number of species, even with mutations and sequencing errors. Opal not only performs up to two orders of magnitude faster than BWA, an alignment-based binning method, but also achieves improved binning accuracy and robustness to sequencing errors. Opal also outperforms models built on traditional k-mer profiles in terms of robustness and accuracy. Finally, we demonstrate that we can effectively use Opal in the "coarse search" stage of a compressive genomics pipeline to identify a much smaller candidate set of taxonomic origins for a subsequent alignment-based method to analyze, thus providing metagenomic binning with high scalability, high accuracy and high resolution.

q-bio.QM↗

arXiv2016

Queue Theory based Response Time Analyses for Geo-Information Processing Chain

Typical characteristics of remote sensing applications are concurrent tasks, such as those found in disaster rapid response. The existing composition approach to geographical information processing service chain, searches for an optimisation solution and is what can be deemed a "selfish" way. This way leads to problems of conflict amongst concurrent tasks and decreases the performance of all service chains. In this study, a non-cooperative game-based mathematical model to analyse the competitive relationships between tasks, is proposed. A best response function is used, to assure each task maintains utility optimisation by considering composition strategies of other tasks and quantifying conflicts between tasks. Based on this, an iterative algorithm that converges to Nash equilibrium is presented, the aim being to provide good convergence and maximise the utilisation of all tasks under concurrent task conditions. Theoretical analyses and experiments showed that the newly proposed method, when compared to existing service composition methods, has better practical utility in all tasks.

cs.DC↗

arXiv2015

Protein secondary structure prediction using deep convolutional neural fields

Protein secondary structure (SS) prediction is important for studying protein structure and function. When only the sequence (profile) information is used as input feature, currently the best predictors can obtain ~80% Q3 accuracy, which has not been improved in the past decade. Here we present DeepCNF (Deep Convolutional Neural Fields) for protein SS prediction. DeepCNF is a Deep Learning extension of Conditional Neural Fields (CNF), which is an integration of Conditional Random Fields (CRF) and shallow neural networks. DeepCNF can model not only complex sequence-structure relationship by a deep hierarchical architecture, but also interdependency between adjacent SS labels, so it is much more powerful than CNF. Experimental results show that DeepCNF can obtain ~84% Q3 accuracy, ~85% SOV score, and ~72% Q8 accuracy, respectively, on the CASP and CAMEO test proteins, greatly outperforming currently popular predictors. As a general framework, DeepCNF can be used to predict other protein structure properties such as contact number, disorder regions, and solvent accessibility.

q-bio.BM↗

arXiv2015

Exact Hybrid Covariance Thresholding for Joint Graphical Lasso

This paper considers the problem of estimating multiple related Gaussian graphical models from a $p$-dimensional dataset consisting of different classes. Our work is based upon the formulation of this problem as group graphical lasso. This paper proposes a novel hybrid covariance thresholding algorithm that can effectively identify zero entries in the precision matrices and split a large joint graphical lasso problem into small subproblems. Our hybrid covariance thresholding method is superior to existing uniform thresholding methods in that our method can split the precision matrix of each individual class using different partition schemes and thus split group graphical lasso into much smaller subproblems, each of which can be solved very fast. In addition, this paper establishes necessary and sufficient conditions for our hybrid covariance thresholding algorithm. The superior performance of our thresholding method is thoroughly analyzed and illustrated by a few experiments on simulated data and real gene expression data.

cs.LG↗

arXiv2015

Bermuda: Bidirectional de novo assembly of transcripts with new insights for handling uneven coverage

Motivation: RNA-seq has made feasible the analysis of a whole set of expressed mRNAs. Mapping-based assembly of RNA-seq reads sometimes is infeasible due to lack of high-quality references. However, de novo assembly is very challenging due to uneven expression levels among transcripts and also the read coverage variation within a single transcript. Existing methods either apply de Bruijn graphs of single-sized k-mers to assemble the full set of transcripts, or conduct multiple runs of assembly, but still apply graphs of single-sized k-mers at each run. However, a single k-mer size is not suitable for all the regions of the transcripts with varied coverage. Contribution: This paper presents a de novo assembler Bermuda with new insights for handling uneven coverage. Opposed to existing methods that use a single k-mer size for all the transcripts in each run of assembly, Bermuda self-adaptively uses a few k-mer sizes to assemble different regions of a single transcript according to their local coverage. As such, Bermuda can deal with uneven expression levels and coverage not only among transcripts, but also within a single transcript. Extensive tests show that Bermuda outperforms popular de novo assemblers in reconstructing unevenly-expressed transcripts with longer length, better contiguity and lower redundancy. Further, Bermuda is computationally efficient with moderate memory consumption.

q-bio.GN↗

arXiv2015

iTreePack: Protein Complex Side-Chain Packing by Dual Decomposition

Protein side-chain packing is a critical component in obtaining the 3D coordinates of a structure and drug discovery. Single-domain protein side-chain packing has been thoroughly studied. A major challenge in generalizing these methods to protein complexes is that they, unlike monomers, often have very large treewidth, and thus algorithms such as TreePack cannot be directly applied. To address this issue, SCWRL4 treats the complex effectively as a monomer, heuristically excluding weak interactions to decrease treewidth; as a result, SCWRL4 generates poor packings on protein interfaces. To date, few side-chain packing methods exist that are specifically designed for protein complexes. In this paper, we introduce a method, iTreePack, which solves the side-chain packing problem for complexes by using a novel combination of dual decomposition and tree decomposition. In particular, iTreePack overcomes the problem of large treewidth by decomposing a protein complex into smaller subgraphs and novelly reformulating the complex side-chain packing problem as a dual relaxation problem; this allows us to solve the side-chain packing of each small subgraph separately using tree-decomposition. A projected subgradient algorithm is applied to enforcing the consistency among the side-chain packings of all the small subgraphs. Computational results demonstrate that our iTreePack program outperforms SCWRL4 on protein complexes. In particular, iTreePack places side-chain atoms much more accurately on very large complexes, which constitute a significant portion of protein-protein interactions. Moreover, the advantage of iTreePack over SCWRL4 increases with respect to the treewidth of a complex. Even for monomeric proteins, iTreePack is much more efficient than SCWRL and slightly more accurate.

q-bio.BM↗