arXiv subjects

Jie Fu

Publications and source records attributed to Jie Fu.

At least 253 records · Page 14Linked to original sources

arXiv2020

An automatic deep learning-based workflow for glioblastoma survival prediction using pre-operative multimodal MR images

We proposed a fully automatic workflow for glioblastoma (GBM) survival prediction using deep learning (DL) methods. 285 glioma (210 GBM, 75 low-grade glioma) patients were included. 163 of the GBM patients had overall survival (OS) data. Every patient had four pre-operative MR scans and manually drawn tumor contours. For automatic tumor segmentation, a 3D convolutional neural network (CNN) was trained and validated using 122 glioma patients. The trained model was applied to the remaining 163 GBM patients to generate tumor contours. The handcrafted and DL-based radiomic features were extracted from auto-contours using explicitly designed algorithms and a pre-trained CNN respectively. 163 GBM patients were randomly split into training (n=122) and testing (n=41) sets for survival analysis. Cox regression models with regularization techniques were trained to construct the handcrafted and DL-based signatures. The prognostic power of the two signatures was evaluated and compared. The 3D CNN achieved an average Dice coefficient of 0.85 across 163 GBM patients for tumor segmentation. The handcrafted signature achieved a C-index of 0.64 (95% CI: 0.55-0.73), while the DL-based signature achieved a C-index of 0.67 (95% CI: 0.57-0.77). Unlike the handcrafted signature, the DL-based signature successfully stratified testing patients into two prognostically distinct groups (p-value<0.01, HR=2.80, 95% CI: 1.26-6.24). The proposed 3D CNN generated accurate GBM tumor contours from four MR images. The DL-based signature resulted in better GBM survival prediction, in terms of higher C-index and significant patient stratification, than the handcrafted signature. The proposed automatic radiomic workflow demonstrated the potential of improving patient stratification and survival prediction in GBM patients.

physics.med-ph↗

arXiv2020

Jacobian Adversarially Regularized Networks for Robustness

Adversarial examples are crafted with imperceptible perturbations with the intent to fool neural networks. Against such attacks, adversarial training and its variants stand as the strongest defense to date. Previous studies have pointed out that robust models that have undergone adversarial training tend to produce more salient and interpretable Jacobian matrices than their non-robust counterparts. A natural question is whether a model trained with an objective to produce salient Jacobian can result in better robustness. This paper answers this question with affirmative empirical results. We propose Jacobian Adversarially Regularized Networks (JARN) as a method to optimize the saliency of a classifier's Jacobian by adversarially regularizing the model's Jacobian to resemble natural training images. Image classifiers trained with JARN show improved robust accuracy compared to standard models on the MNIST, SVHN and CIFAR-10 datasets, uncovering a new angle to boost robustness without using adversarial training examples.

cs.CV↗

arXiv2019

Revision in Continuous Space: Unsupervised Text Style Transfer without Adversarial Learning

Typical methods for unsupervised text style transfer often rely on two key ingredients: 1) seeking the explicit disentanglement of the content and the attributes, and 2) troublesome adversarial learning. In this paper, we show that neither of these components is indispensable. We propose a new framework that utilizes the gradients to revise the sentence in a continuous space during inference to achieve text style transfer. Our method consists of three key components: a variational auto-encoder (VAE), some attribute predictors (one for each attribute), and a content predictor. The VAE and the two types of predictors enable us to perform gradient-based optimization in the continuous space, which is mapped from sentences in a discrete space, to find the representation of a target sentence with the desired attributes and preserved content. Moreover, the proposed method naturally has the ability to simultaneously manipulate multiple fine-grained attributes, such as sentence length and the presence of specific words, when performing text style transfer tasks. Compared with previous adversarial learning based methods, the proposed method is more interpretable, controllable and easier to train. Extensive experimental studies on three popular text style transfer tasks show that the proposed method significantly outperforms five state-of-the-art methods.

cs.CL↗

arXiv2019

Topological Approximate Dynamic Programming under Temporal Logic Constraints

In this paper, we develop a Topological Approximate Dynamic Programming (TADP) method for planningin stochastic systems modeled as Markov Decision Processesto maximize the probability of satisfying high-level systemspecifications expressed in Linear Temporal Logic (LTL). Ourmethod includes two steps: First, we propose to decompose theplanning problem into a sequence of sub-problems based on thetopological property of the task automaton which is translatedfrom the LTL constraints. Second, we extend a model-freeapproximate dynamic programming method for value iterationto solve, in an order reverse to a causal dependency of valuefunctions, one for each state in the task automaton. Particularly,we show that the complexity of the TADP does not growpolynomially with the size of the product Markov DecisionProcess (MDP). The correctness and efficiency of the algorithmare demonstrated using a robotic motion planning example.

math.OC↗

arXiv2019

Learning Sparse Mixture of Experts for Visual Question Answering

There has been a rapid progress in the task of Visual Question Answering with improved model architectures. Unfortunately, these models are usually computationally intensive due to their sheer size which poses a serious challenge for deployment. We aim to tackle this issue for the specific task of Visual Question Answering (VQA). A Convolutional Neural Network (CNN) is an integral part of the visual processing pipeline of a VQA model (assuming the CNN is trained along with entire VQA model). In this project, we propose an efficient and modular neural architecture for the VQA task with focus on the CNN module. Our experiments demonstrate that a sparsely activated CNN based VQA model achieves comparable performance to a standard CNN based VQA model architecture.

cs.LG↗

arXiv2019

Deep Learning-based Radiomic Features for Improving Neoadjuvant Chemoradiation Response Prediction in Locally Advanced Rectal Cancer

Radiomic features achieve promising results in cancer diagnosis, treatment response prediction, and survival prediction. Our goal is to compare the handcrafted (explicitly designed) and deep learning (DL)-based radiomic features extracted from pre-treatment diffusion-weighted magnetic resonance images (DWIs) for predicting neoadjuvant chemoradiation treatment (nCRT) response in patients with locally advanced rectal cancer (LARC). 43 patients receiving nCRT were included. All patients underwent DWIs before nCRT and total mesorectal excision surgery 6-12 weeks after completion of nCRT. Gross tumor volume (GTV) contours were drawn by an experienced radiation oncologist on DWIs. The patient-cohort was split into the responder group (n=22) and the non-responder group (n=21) based on the post-nCRT response assessed by postoperative pathology, MRI or colonoscopy. Handcrafted and DL-based features were extracted from the apparent diffusion coefficient (ADC) map of the DWI using conventional computer-aided diagnosis methods and a pre-trained convolution neural network, respectively. Least absolute shrinkage and selection operator (LASSO)-logistic regression models were constructed using extracted features for predicting treatment response. The model performance was evaluated with repeated 20 times stratified 4-fold cross-validation using receiver operating characteristic (ROC) curves and compared using the corrected resampled t-test. The model built with handcrafted features achieved the mean area under the ROC curve (AUC) of 0.64, while the one built with DL-based features yielded the mean AUC of 0.73. The corrected resampled t-test on AUC showed P-value < 0.05. DL-based features extracted from pre-treatment DWIs achieved significantly better classification performance compared with handcrafted features for predicting nCRT response in patients with LARC.

physics.med-ph↗

arXiv2019

Exploring Domain Shift in Extractive Text Summarization

Although domain shift has been well explored in many NLP applications, it still has received little attention in the domain of extractive text summarization. As a result, the model is under-utilizing the nature of the training data due to ignoring the difference in the distribution of training sets and shows poor generalization on the unseen domain. With the above limitation in mind, in this paper, we first extend the conventional definition of the domain from categories into data sources for the text summarization task. Then we re-purpose a multi-domain summarization dataset and verify how the gap between different domains influences the performance of neural summarization models. Furthermore, we investigate four learning strategies and examine their abilities to deal with the domain shift problem. Experimental results on three different settings show their different characteristics in our new testbed. Our source code including \textit{BERT-based}, \textit{meta-learning} methods for multi-domain summarization learning and the re-purposed dataset \textsc{Multi-SUM} will be available on our project: \url{http://pfliu.com/TransferSum/}.

cs.CL↗

arXiv2019

Interactive Language Learning by Question Answering

Humans observe and interact with the world to acquire knowledge. However, most existing machine reading comprehension (MRC) tasks miss the interactive, information-seeking component of comprehension. Such tasks present models with static documents that contain all necessary information, usually concentrated in a single short substring. Thus, models can achieve strong performance through simple word- and phrase-based pattern matching. We address this problem by formulating a novel text-based question answering task: Question Answering with Interactive Text (QAit). In QAit, an agent must interact with a partially observable text-based environment to gather information required to answer questions. QAit poses questions about the existence, location, and attributes of objects found in the environment. The data is built using a text-based game generator that defines the underlying dynamics of interaction with the environment. We propose and evaluate a set of baseline models for the QAit task that includes deep reinforcement learning agents. Experiments show that the task presents a major challenge for machine reading systems, while humans solve it with relative ease.

cs.CL↗

arXiv2019

Generation of abdominal synthetic CTs from 0.35T MR images using generative adversarial networks for MR-only liver radiotherapy

Electron density maps must be accurately estimated to achieve valid dose calculation in MR-only radiotherapy. The goal of this study is to assess whether two deep learning models, the conditional generative adversarial network (cGAN) and the cycle-consistent generative adversarial network (cycleGAN), can generate accurate abdominal synthetic CT (sCT) images from 0.35T MR images for MR-only liver radiotherapy. A retrospective study was performed using CT images and 0.35T MR images of 12 patients with liver (n=8) and non-liver abdominal (n=4) cancer. CT images were deformably registered to the corresponding MR images to generate deformed CT (dCT) images for treatment planning. Both cGAN and cycleGAN were trained using MR and dCT transverse slices. Four-fold cross-validation testing was conducted to generate sCT images for all patients. The HU prediction accuracy was evaluated by voxel-wise similarity metric between each dCT and sCT image for all 12 patients. dCT-based and sCT-based dose distributions were compared using gamma and dose-volume histogram (DVH) metric analysis for 8 liver patients. sCTcycleGAN achieved the average mean absolute error (MAE) of 94.1 HU, while sCTcGAN achieved 89.8 HU. In both models, the average gamma passing rates within all volumes of interest were higher than 95% using a 2%, 2 mm criterion, and 99% using a 3%, 3 mm criterion. The average differences in the mean dose and DVH metrics were within +/-0.6% for the planning target volume and within +/-0.15% for evaluated organs in both models. Results demonstrated that abdominal sCT images generated by both cGAN and cycleGAN achieved accurate dose calculation for 8 liver radiotherapy plans. sCTcGAN images had smaller average MAE and achieved better dose calculation accuracy than sCTcyleGAN images. More abdominal patients will be enrolled in the future to further evaluate two models.

physics.med-ph↗

arXiv2019

Compositional planning in Markov decision processes: Temporal abstraction meets generalized logic composition

In hierarchical planning for Markov decision processes (MDPs), temporal abstraction allows planning with macro-actions that take place at different time scale in form of sequential composition. In this paper, we propose a novel approach to compositional reasoning and hierarchical planning for MDPs under temporal logic constraints. In addition to sequential composition, we introduce a composition of policies based on generalized logic composition: Given sub-policies for sub-tasks and a new task expressed as logic compositions of subtasks, a semi-optimal policy, which is optimal in planning with only sub-policies, can be obtained by simply composing sub-polices. Thus, a synthesis algorithm is developed to compute optimal policies efficiently by planning with primitive actions, policies for sub-tasks, and the compositions of sub-policies, for maximizing the probability of satisfying temporal logic specifications. We demonstrate the correctness and efficiency of the proposed method in stochastic planning examples with a single agent and multiple task specifications.

math.OC↗

arXiv2019

Conditional Computation for Continual Learning

Catastrophic forgetting of connectionist neural networks is caused by the global sharing of parameters among all training examples. In this study, we analyze parameter sharing under the conditional computation framework where the parameters of a neural network are conditioned on each input example. At one extreme, if each input example uses a disjoint set of parameters, there is no sharing of parameters thus no catastrophic forgetting. At the other extreme, if the parameters are the same for every example, it reduces to the conventional neural network. We then introduce a clipped version of maxout networks which lies in the middle, i.e. parameters are shared partially among examples. Based on the parameter sharing analysis, we can locate a limited set of examples that are interfered when learning a new example. We propose to perform rehearsal on this set to prevent forgetting, which is termed as conditional rehearsal. Finally, we demonstrate the effectiveness of the proposed method in an online non-stationary setup, where updates are made after each new example and the distribution of the received example shifts over time.

cs.LG↗

arXiv2019

Opportunistic Synthesis in Reactive Games under Information Asymmetry

Reactive synthesis is a class of methods to construct a provably-correct control system, referred to as a robot, with respect to a temporal logic specification in the presence of a dynamic and uncontrollable environment. This is achieved by modeling the interaction between the robot and its environment as a two-player zero-sum game. However, existing reactive synthesis methods assume both players to have complete information, which is not the case in many strategic interactions. In this paper, we use a variant of hypergames to model the interaction between the robot and its environment; which has incomplete information about the specification of the robot. This model allows us to identify a subset of game states from where the robot can leverage the asymmetrical information to achieve a better outcome, which is not possible if both players have symmetrical and complete information. We then introduce a novel method of opportunistic synthesis by defining a Markov Decision Process (MDP) using the hypergame under temporal logic specifications. When the environment plays some stochastic strategy in its perceived sure-winning and sure-losing regions of the game, we show that by following the opportunistic strategy, the robot is ensured to only improve the outcome of the game - measured by satisfaction of sub-specifications - whenever an opportunity becomes available. We demonstrate the correctness and optimality of this method using a robot motion planning example in the presence of an adversary.

cs.FL↗

arXiv2019

Lightweight and Efficient Neural Natural Language Processing with Quaternion Networks

Many state-of-the-art neural models for NLP are heavily parameterized and thus memory inefficient. This paper proposes a series of lightweight and memory efficient neural architectures for a potpourri of natural language processing (NLP) tasks. To this end, our models exploit computation using Quaternion algebra and hypercomplex spaces, enabling not only expressive inter-component interactions but also significantly ($75\%$) reduced parameter size due to lesser degrees of freedom in the Hamilton product. We propose Quaternion variants of models, giving rise to new architectures such as the Quaternion attention Model and Quaternion Transformer. Extensive experiments on a battery of NLP tasks demonstrates the utility of proposed Quaternion-inspired models, enabling up to $75\%$ reduction in parameter size without significant loss in performance.

cs.CL↗

arXiv2019

Structure Learning for Neural Module Networks

Neural Module Networks, originally proposed for the task of visual question answering, are a class of neural network architectures that involve human-specified neural modules, each designed for a specific form of reasoning. In current formulations of such networks only the parameters of the neural modules and/or the order of their execution is learned. In this work, we further expand this approach and also learn the underlying internal structure of modules in terms of the ordering and combination of simple and elementary arithmetic operators. Our results show that one is indeed able to simultaneously learn both internal module structure and module sequencing without extra supervisory signals for module execution sequencing. With this approach, we report performance comparable to models using hand-designed modules.

cs.LG↗

arXiv2019

Simple and Effective Curriculum Pointer-Generator Networks for Reading Comprehension over Long Narratives

This paper tackles the problem of reading comprehension over long narratives where documents easily span over thousands of tokens. We propose a curriculum learning (CL) based Pointer-Generator framework for reading/sampling over large documents, enabling diverse training of the neural model based on the notion of alternating contextual difficulty. This can be interpreted as a form of domain randomization and/or generative pretraining during training. To this end, the usage of the Pointer-Generator softens the requirement of having the answer within the context, enabling us to construct diverse training samples for learning. Additionally, we propose a new Introspective Alignment Layer (IAL), which reasons over decomposed alignments using block-based self-attention. We evaluate our proposed method on the NarrativeQA reading comprehension benchmark, achieving state-of-the-art performance, improving existing baselines by $51\%$ relative improvement on BLEU-4 and $17\%$ relative improvement on Rouge-L. Extensive ablations confirm the effectiveness of our proposed IAL and CL components.

cs.CL↗

arXiv2019

TIGS: An Inference Algorithm for Text Infilling with Gradient Search

Text infilling is defined as a task for filling in the missing part of a sentence or paragraph, which is suitable for many real-world natural language generation scenarios. However, given a well-trained sequential generative model, generating missing symbols conditioned on the context is challenging for existing greedy approximate inference algorithms. In this paper, we propose an iterative inference algorithm based on gradient search, which is the first inference algorithm that can be broadly applied to any neural sequence generative models for text infilling tasks. We compare the proposed method with strong baselines on three text infilling tasks with various mask ratios and different mask strategies. The results show that our proposed method is effective and efficient for fill-in-the-blank tasks, consistently outperforming all baselines.

cs.CL↗

arXiv2019

A Validated Physical Model For Real-Time Simulation of Soft Robotic Snakes

In this work we present a framework that is capable of accurately representing soft robotic actuators in a multiphysics environment in real-time. We propose a constraint-based dynamics model of a 1-dimensional pneumatic soft actuator that accounts for internal pressure forces, as well as the effect of actuator latency and damping under inflation and deflation and demonstrate its accuracy a full soft robotic snake with the composition of multiple 1D actuators. We verify our model's accuracy in static deformation and dynamic locomotion open-loop control experiments. To achieve real-time performance we leverage the parallel computation power of GPUs to allow interactive control and feedback.

cs.RO↗

arXiv2019

Graph Neural Networks with Generated Parameters for Relation Extraction

Recently, progress has been made towards improving relational reasoning in machine learning field. Among existing models, graph neural networks (GNNs) is one of the most effective approaches for multi-hop relational reasoning. In fact, multi-hop relational reasoning is indispensable in many natural language processing tasks such as relation extraction. In this paper, we propose to generate the parameters of graph neural networks (GP-GNNs) according to natural language sentences, which enables GNNs to process relational reasoning on unstructured text inputs. We verify GP-GNNs in relation extraction from text. Experimental results on a human-annotated dataset and two distantly supervised datasets show that our model achieves significant improvements compared to baselines. We also perform a qualitative analysis to demonstrate that our model could discover more accurate relations by multi-hop relational reasoning.

cs.CL↗