arXiv subjects

Liwei Wang

Publications and source records attributed to Liwei Wang.

At least 253 records · Page 14Linked to original sources

arXiv2018

Transferrable Feature and Projection Learning with Class Hierarchy for Zero-Shot Learning

Zero-shot learning (ZSL) aims to transfer knowledge from seen classes to unseen ones so that the latter can be recognised without any training samples. This is made possible by learning a projection function between a feature space and a semantic space (e.g. attribute space). Considering the seen and unseen classes as two domains, a big domain gap often exists which challenges ZSL. Inspired by the fact that an unseen class is not exactly `unseen' if it belongs to the same superclass as a seen class, we propose a novel inductive ZSL model that leverages superclasses as the bridge between seen and unseen classes to narrow the domain gap. Specifically, we first build a class hierarchy of multiple superclass layers and a single class layer, where the superclasses are automatically generated by data-driven clustering over the semantic representations of all seen and unseen class names. We then exploit the superclasses from the class hierarchy to tackle the domain gap challenge in two aspects: deep feature learning and projection function learning. First, to narrow the domain gap in the feature space, we integrate a recurrent neural network (RNN) defined with the superclasses into a convolutional neural network (CNN), in order to enforce the superclass hierarchy. Second, to further learn a transferrable projection function for ZSL, a novel projection function learning method is proposed by exploiting the superclasses to align the two domains. Importantly, our transferrable feature and projection learning methods can be easily extended to a closely related task -- few-shot learning (FSL). Extensive experiments show that the proposed model significantly outperforms the state-of-the-art alternatives in both ZSL and FSL tasks.

cs.CV↗

arXiv2018

Learning to Navigate for Fine-grained Classification

Fine-grained classification is challenging due to the difficulty of finding discriminative features. Finding those subtle traits that fully characterize the object is not straightforward. To handle this circumstance, we propose a novel self-supervision mechanism to effectively localize informative regions without the need of bounding-box/part annotations. Our model, termed NTS-Net for Navigator-Teacher-Scrutinizer Network, consists of a Navigator agent, a Teacher agent and a Scrutinizer agent. In consideration of intrinsic consistency between informativeness of the regions and their probability being ground-truth class, we design a novel training paradigm, which enables Navigator to detect most informative regions under the guidance from Teacher. After that, the Scrutinizer scrutinizes the proposed regions from Navigator and makes predictions. Our model can be viewed as a multi-agent cooperation, wherein agents benefit from each other, and make progress together. NTS-Net can be trained end-to-end, while provides accurate fine-grained classification predictions as well as highly informative regions during inference. We achieve state-of-the-art performance in extensive benchmark datasets.

cs.CV↗

arXiv2018

MedSTS: A Resource for Clinical Semantic Textual Similarity

The wide adoption of electronic health records (EHRs) has enabled a wide range of applications leveraging EHR data. However, the meaningful use of EHR data largely depends on our ability to efficiently extract and consolidate information embedded in clinical text where natural language processing (NLP) techniques are essential. Semantic textual similarity (STS) that measures the semantic similarity between text snippets plays a significant role in many NLP applications. In the general NLP domain, STS shared tasks have made available a huge collection of text snippet pairs with manual annotations in various domains. In the clinical domain, STS can enable us to detect and eliminate redundant information that may lead to a reduction in cognitive burden and an improvement in the clinical decision-making process. This paper elaborates our efforts to assemble a resource for STS in the medical domain, MedSTS. It consists of a total of 174,629 sentence pairs gathered from a clinical corpus at Mayo Clinic. A subset of MedSTS (MedSTS_ann) containing 1,068 sentence pairs was annotated by two medical experts with semantic similarity scores of 0-5 (low to high similarity). We further analyzed the medical concepts in the MedSTS corpus, and tested four STS systems on the MedSTS_ann corpus. In the future, we will organize a shared task by releasing the MedSTS_ann corpus to motivate the community to tackle the real world clinical problems.

cs.IR↗

arXiv2018

A Comparison of Word Embeddings for the Biomedical Natural Language Processing

Word embeddings have been widely used in biomedical Natural Language Processing (NLP) applications as they provide vector representations of words capturing the semantic properties of words and the linguistic relationship between words. Many biomedical applications use different textual resources (e.g., Wikipedia and biomedical articles) to train word embeddings and apply these word embeddings to downstream biomedical applications. However, there has been little work on evaluating the word embeddings trained from these resources.In this study, we provide an empirical evaluation of word embeddings trained from four different resources, namely clinical notes, biomedical publications, Wikipedia, and news. We performed the evaluation qualitatively and quantitatively. For the qualitative evaluation, we manually inspected five most similar medical words to a given set of target medical words, and then analyzed word embeddings through the visualization of those word embeddings. For the quantitative evaluation, we conducted both intrinsic and extrinsic evaluation. Based on the evaluation results, we can draw the following conclusions. First, the word embeddings trained on clinical notes and biomedical publications can capture the semantics of medical terms better, and find more relevant similar medical terms, and are closer to human experts' judgments, compared to these trained on Wikipedia and news. Second, there does not exist a consistent global ranking of word embedding quality for downstream biomedical NLP applications. However, adding word embeddings as extra features will improve results on most downstream tasks. Finally, the word embeddings trained on biomedical domain corpora do not necessarily have better performance than those trained on other general domain corpora for any downstream biomedical NLP tasks.

cs.IR↗

arXiv2018

Towards Binary-Valued Gates for Robust LSTM Training

Long Short-Term Memory (LSTM) is one of the most widely used recurrent structures in sequence modeling. It aims to use gates to control information flow (e.g., whether to skip some information or not) in the recurrent computations, although its practical implementation based on soft gates only partially achieves this goal. In this paper, we propose a new way for LSTM training, which pushes the output values of the gates towards 0 or 1. By doing so, we can better control the information flow: the gates are mostly open or closed, instead of in a middle state, which makes the results more interpretable. Empirical studies show that (1) Although it seems that we restrict the model capacity, there is no performance drop: we achieve better or comparable performances due to its better generalization ability; (2) The outputs of gates are not sensitive to their inputs: we can easily compress the LSTM unit in multiple ways, e.g., low-rank approximation and low-precision approximation. The compressed models are even better than the baseline models without compression.

cs.LG↗

arXiv2018

FPGA Based Pico-second Time Measurement System for a DIRC-like TOF Detector

A prototype of DIRC-like Time-of-Flight detector (DTOF), including a pico-second time measurement electronics, is developed and tested preliminarily. The basic structure of DTOF is composed of a fused silica radiator connected to fast micro-channel plate PMTs (MCP-PMT), and readout by a dedicated FPGA (Field Programmable Gate Array) based front-end electronics. The full electronics chain consists of a programmable differential amplifier, a dual-threshold differential discriminator, and a timestamp Time-to-Digital convertor. By splitting a MCP-PMT output signal into two identical electronics chains, the coincidence time resolution (CTR) of pure electronics was measured as 5.6 ps. By the beam test in H4 (150GeV/c, Muon) at CERN, the intrinsic CTR of the whole detector prototype reaches 15.0 ps without using time-amplitude correction. The test results demonstrate that the FPGA based front-end electronics could achieve an excellent time performance for TOF detectors. It is very compact, cost effective with a high multi-channel capacity and short measurement dead time, which is very suitable for practical applications of large-scale high performance TOF detectors in particle physics spectrometer.

physics.ins-det↗

arXiv2018

Study on Timing Performance of a Readout Circuit for SiPM

In recent years, SiPM photoelectric devices have drawn much attention in the domain of time-of-flight-based positron emission tomography (TOF-PET). Using them to construct PET detectors with excellent coincidence time resolution (CTR) is always one of research focus. In this paper, a SiPM readout pre-amplifier based on common-base current amplifier structure followed by a Pole-Zero (PZ) compensation network is constructed, and the main factors that affect the timing performance of the PET detector are investigated. By experimental measurement, we found that the CTR is heavily related to the bandwidth of the amplifier, bias voltage of SiPM, comparator threshold, and PZ network parameter. The test setup has two detectors, one with LYSO crystal (3 mm 3 mm 10 mm) coupled with a Hamamatsu SiPM (S12642-0404), and the other with LaBr3 coupled to a PMT-R9800. After the optimization of the readout circuit with related factors, the CTR between the two detectors is measured as 266ps FWHM. The test result is a helpful guideline for the readout ASIC chip design in our next step.

physics.ins-det↗

arXiv2018

An FPGA Based Fast Linear Discharge Method for Nuclear Pulse Digitization

Inspired by Wilkinson ADC method, we implement a fast linear discharge method based on FPGA to digitize nuclear pulse signal. In this scheme, we use a constant current source to discharge the charge on capacitor which is integrated by the input current pulse so as to convert the amplitude of the input nuclear pulse into time width linearly. Thanks to the high precision of TDC measurement that we have achieved in FPGA, we can increase the current value of the discharge to make the discharge time short, so as to obtain a small measurement of dead time. We has realized a single channel fast linear discharge circuit which contains only one dual supply amplifier, two resistors and one capacitor. The rest part can be implemented in an FPGA (Field Programmable Gate Array). Leakage current from the sensor would cause the base line drifting slowly, which can influence the measuring precision. Our method solve this problem without losing the linearity of measurement. We have built the circuit and experimental setup for evaluation. Using them to measure energy spectrums of PET detectors of PMT coupled with LYSO and LaBr3 crystal, the energy resolution is 12.67% and 5.17% respectively. The test results show that our circuit is rather simple, stable and conducive for multi-channel integration.

physics.ins-det↗

arXiv2018

Learning Two-Branch Neural Networks for Image-Text Matching Tasks

Image-language matching tasks have recently attracted a lot of attention in the computer vision field. These tasks include image-sentence matching, i.e., given an image query, retrieving relevant sentences and vice versa, and region-phrase matching or visual grounding, i.e., matching a phrase to relevant regions. This paper investigates two-branch neural networks for learning the similarity between these two data modalities. We propose two network structures that produce different output representations. The first one, referred to as an embedding network, learns an explicit shared latent embedding space with a maximum-margin ranking loss and novel neighborhood constraints. Compared to standard triplet sampling, we perform improved neighborhood sampling that takes neighborhood information into consideration while constructing mini-batches. The second network structure, referred to as a similarity network, fuses the two branches via element-wise product and is trained with regression loss to directly predict a similarity score. Extensive experiments show that our networks achieve high accuracies for phrase localization on the Flickr30K Entities dataset and for bi-directional image-sentence retrieval on Flickr30K and MSCOCO datasets.

cs.CV↗

arXiv2018

A Deep Representation Empowered Distant Supervision Paradigm for Clinical Information Extraction

Objective: To automatically create large labeled training datasets and reduce the efforts of feature engineering for training accurate machine learning models for clinical information extraction. Materials and Methods: We propose a distant supervision paradigm empowered by deep representation for extracting information from clinical text. In this paradigm, the rule-based NLP algorithms are utilized to generate weak labels and create large training datasets automatically. Additionally, we use pre-trained word embeddings as deep representation to eliminate the need of task-specific feature engineering for machine learning. We evaluated the effectiveness of the proposed paradigm on two clinical information extraction tasks: smoking status extraction and proximal femur (hip) fracture extraction. We tested three prevalent machine learning models, namely, Convolutional Neural Networks (CNN), Support Vector Machine (SVM), and Random Forrest (RF). Results: The results indicate that CNN is the best fit to the proposed distant supervision paradigm. It outperforms the rule-based NLP algorithms given large datasets by capturing additional extraction patterns. We also verified the advantage of word embedding feature representation in the paradigm over term frequency-inverse document frequency (tf-idf) and topic modeling representations. Discussion: In the clinical domain, the limited amount of labeled data is always a bottleneck for applying machine learning. Additionally, the performance of machine learning approaches highly depends on task-specific feature engineering. The proposed paradigm could alleviate those problems by leveraging rule-based NLP algorithms to automatically assign weak labels and eliminating the need of task-specific feature engineering using word embedding feature representation.

cs.IR↗

arXiv2018

Learning Region Features for Object Detection

While most steps in the modern object detection methods are learnable, the region feature extraction step remains largely hand-crafted, featured by RoI pooling methods. This work proposes a general viewpoint that unifies existing region feature extraction methods and a novel method that is end-to-end learnable. The proposed method removes most heuristic choices and outperforms its RoI pooling counterparts. It moves further towards fully learnable object detection.

cs.CV↗

arXiv2018

Randomness in Deconvolutional Networks for Visual Representation

Toward a deeper understanding on the inner work of deep neural networks, we investigate CNN (convolutional neural network) using DCN (deconvolutional network) and randomization technique, and gain new insights for the intrinsic property of this network architecture. For the random representations of an untrained CNN, we train the corresponding DCN to reconstruct the input images. Compared with the image inversion on pre-trained CNN, our training converges faster and the yielding network exhibits higher quality for image reconstruction. It indicates there is rich information encoded in the random features; the pre-trained CNN may discard information irrelevant for classification and encode relevant features in a way favorable for classification but harder for reconstruction. We further explore the property of the overall random CNN-DCN architecture. Surprisingly, images can be inverted with satisfactory quality. Extensive empirical evidence as well as theoretical analysis are provided.

cs.CV↗

arXiv2017

Diverse and Accurate Image Description Using a Variational Auto-Encoder with an Additive Gaussian Encoding Space

This paper explores image caption generation using conditional variational auto-encoders (CVAEs). Standard CVAEs with a fixed Gaussian prior yield descriptions with too little variability. Instead, we propose two models that explicitly structure the latent space around $K$ components corresponding to different types of image content, and combine components to create priors for images that contain multiple types of content simultaneously (e.g., several kinds of objects). Our first model uses a Gaussian Mixture model (GMM) prior, while the second one defines a novel Additive Gaussian (AG) prior that linearly combines component means. We show that both models produce captions that are more diverse and more accurate than a strong LSTM baseline or a "vanilla" CVAE with a fixed Gaussian prior, with AG-CVAE showing particular promise.

cs.CV↗

arXiv2017

The Expressive Power of Neural Networks: A View from the Width

The expressive power of neural networks is important for understanding deep learning. Most existing works consider this problem from the view of the depth of a network. In this paper, we study how width affects the expressiveness of neural networks. Classical results state that depth-bounded (e.g. depth-$2$) networks with suitable activation functions are universal approximators. We show a universal approximation theorem for width-bounded ReLU networks: width-$(n+4)$ ReLU networks, where $n$ is the input dimension, are universal approximators. Moreover, except for a measure zero set, all functions cannot be approximated by width-$n$ ReLU networks, which exhibits a phase transition. Several recent works demonstrate the benefits of depth by proving the depth-efficiency of neural networks. That is, there are classes of deep networks which cannot be realized by any shallow network whose size is no more than an exponential bound. Here we pose the dual question on the width-efficiency of ReLU networks: Are there wide networks that cannot be realized by narrow networks whose size is not substantially larger? We show that there exist classes of wide networks which cannot be realized by any narrow network whose depth is no more than a polynomial bound. On the other hand, we demonstrate by extensive experiments that narrow networks whose size exceed the polynomial bound by a constant factor can approximate wide and shallow network with high accuracy. Our results provide more comprehensive evidence that depth is more effective than width for the expressiveness of ReLU networks.

cs.LG↗

arXiv2017

Accurate Pulmonary Nodule Detection in Computed Tomography Images Using Deep Convolutional Neural Networks

Early detection of pulmonary cancer is the most promising way to enhance a patient's chance for survival. Accurate pulmonary nodule detection in computed tomography (CT) images is a crucial step in diagnosing pulmonary cancer. In this paper, inspired by the successful use of deep convolutional neural networks (DCNNs) in natural image recognition, we propose a novel pulmonary nodule detection approach based on DCNNs. We first introduce a deconvolutional structure to Faster Region-based Convolutional Neural Network (Faster R-CNN) for candidate detection on axial slices. Then, a three-dimensional DCNN is presented for the subsequent false positive reduction. Experimental results of the LUng Nodule Analysis 2016 (LUNA16) Challenge demonstrate the superior detection performance of the proposed approach on nodule detection(average FROC-score of 0.891, ranking the 1st place over all submitted results).

cs.CV↗

arXiv2017

Generalization Bounds of SGLD for Non-convex Learning: Two Theoretical Viewpoints

Algorithm-dependent generalization error bounds are central to statistical learning theory. A learning algorithm may use a large hypothesis space, but the limited number of iterations controls its model capacity and generalization error. The impacts of stochastic gradient methods on generalization error for non-convex learning problems not only have important theoretical consequences, but are also critical to generalization errors of deep learning. In this paper, we study the generalization errors of Stochastic Gradient Langevin Dynamics (SGLD) with non-convex objectives. Two theories are proposed with non-asymptotic discrete-time analysis, using Stability and PAC-Bayesian results respectively. The stability-based theory obtains a bound of $O\left(\frac{1}{n}L\sqrt{βT_k}\right)$, where $L$ is uniform Lipschitz parameter, $β$ is inverse temperature, and $T_k$ is aggregated step sizes. For PAC-Bayesian theory, though the bound has a slower $O(1/\sqrt{n})$ rate, the contribution of each step is shown with an exponentially decaying factor by imposing $\ell^2$ regularization, and the uniform Lipschitz constant is also replaced by actual norms of gradients along trajectory. Our bounds have no implicit dependence on dimensions, norms or other capacity measures of parameter, which elegantly characterizes the phenomenon of "Fast Training Guarantees Generalization" in non-convex settings. This is the first algorithm-dependent result with reasonable dependence on aggregated step sizes for non-convex learning, and has important implications to statistical learning aspects of stochastic gradient methods in complicated models such as deep learning.

cs.LG↗

arXiv2017

Zero-Shot Fine-Grained Classification by Deep Feature Learning with Semantics

Fine-grained image classification, which aims to distinguish images with subtle distinctions, is a challenging task due to two main issues: lack of sufficient training data for every class and difficulty in learning discriminative features for representation. In this paper, to address the two issues, we propose a two-phase framework for recognizing images from unseen fine-grained classes, i.e. zero-shot fine-grained classification. In the first feature learning phase, we finetune deep convolutional neural networks using hierarchical semantic structure among fine-grained classes to extract discriminative deep visual features. Meanwhile, a domain adaptation structure is induced into deep convolutional neural networks to avoid domain shift from training data to test data. In the second label inference phase, a semantic directed graph is constructed over attributes of fine-grained classes. Based on this graph, we develop a label propagation algorithm to infer the labels of images in the unseen classes. Experimental results on two benchmark datasets demonstrate that our model outperforms the state-of-the-art zero-shot learning models. In addition, the features obtained by our feature learning model also yield significant gains when they are used by other zero-shot learning models, which shows the flexility of our model in zero-shot fine-grained classification.

cs.CV↗

arXiv2017

Collect at Once, Use Effectively: Making Non-interactive Locally Private Learning Possible

Non-interactive Local Differential Privacy (LDP) requires data analysts to collect data from users through noisy channel at once. In this paper, we extend the frontiers of Non-interactive LDP learning and estimation from several aspects. For learning with smooth generalized linear losses, we propose an approximate stochastic gradient oracle estimated from non-interactive LDP channel, using Chebyshev expansion. Combined with inexact gradient methods, we obtain an efficient algorithm with quasi-polynomial sample complexity bound. For the high-dimensional world, we discover that under $\ell_2$-norm assumption on data points, high-dimensional sparse linear regression and mean estimation can be achieved with logarithmic dependence on dimension, using random projection and approximate recovery. We also extend our methods to Kernel Ridge Regression. Our work is the first one that makes learning and estimation possible for a broad range of learning tasks under non-interactive LDP model.

cs.LG↗