arXiv subjects

Jie Fu

Publications and source records attributed to Jie Fu.

At least 235 records · Page 13Linked to original sources

arXiv2021

Beyond Fully-Connected Layers with Quaternions: Parameterization of Hypercomplex Multiplications with $1/n$ Parameters

Recent works have demonstrated reasonable success of representation learning in hypercomplex space. Specifically, "fully-connected layers with Quaternions" (4D hypercomplex numbers), which replace real-valued matrix multiplications in fully-connected layers with Hamilton products of Quaternions, both enjoy parameter savings with only 1/4 learnable parameters and achieve comparable performance in various applications. However, one key caveat is that hypercomplex space only exists at very few predefined dimensions (4D, 8D, and 16D). This restricts the flexibility of models that leverage hypercomplex multiplications. To this end, we propose parameterizing hypercomplex multiplications, allowing models to learn multiplication rules from data regardless of whether such rules are predefined. As a result, our method not only subsumes the Hamilton product, but also learns to operate on any arbitrary nD hypercomplex space, providing more architectural flexibility using arbitrarily $1/n$ learnable parameters compared with the fully-connected layer counterpart. Experiments of applications to the LSTM and Transformer models on natural language inference, machine translation, text style transfer, and subject verb agreement demonstrate architectural flexibility and effectiveness of the proposed approach.

cs.LG↗

arXiv2020

Attention-Based Planning with Active Perception

Attention control is a key cognitive ability for humans to select information relevant to the current task. This paper develops a computational model of attention and an algorithm for attention-based probabilistic planning in Markov decision processes. In attention-based planning, the robot decides to be in different attention modes. An attention mode corresponds to a subset of state variables monitored by the robot. By switching between different attention modes, the robot actively perceives task-relevant information to reduce the cost of information acquisition and processing, while achieving near-optimal task performance. Though planning with attention-based active perception inevitably introduces partial observations, a partially observable MDP formulation makes the problem computational expensive to solve. Instead, our proposed method employs a hierarchical planning framework in which the robot determines what to pay attention to and for how long the attention should be sustained before shifting to other information sources. During the attention sustaining phase, the robot carries out a sub-policy, computed from an abstraction of the original MDP given the current attention. We use an example where a robot is tasked to capture a set of intruders in a stochastic gridworld. The experimental results show that the proposed method enables information- and computation-efficient optimal planning in stochastic environments.

cs.RO↗

arXiv2020

Semantic SLAM with Autonomous Object-Level Data Association

It is often desirable to capture and map semantic information of an environment during simultaneous localization and mapping (SLAM). Such semantic information can enable a robot to better distinguish places with similar low-level geometric and visual features and perform high-level tasks that use semantic information about objects to be manipulated and environments to be navigated. While semantic SLAM has gained increasing attention, there is little research on semanticlevel data association based on semantic objects, i.e., object-level data association. In this paper, we propose a novel object-level data association algorithm based on bag of words algorithm, formulated as a maximum weighted bipartite matching problem. With object-level data association solved, we develop a quadratic-programming-based semantic object initialization scheme using dual quadric and introduce additional constraints to improve the success rate of object initialization. The integrated semantic-level SLAM system can achieve high-accuracy object-level data association and real-time semantic mapping as demonstrated in the experiments. The online semantic map building and semantic-level localization capabilities facilitate semantic-level mapping and task planning in a priori unknown environment.

cs.RO↗

arXiv2020

Diverse, Controllable, and Keyphrase-Aware: A Corpus and Method for News Multi-Headline Generation

News headline generation aims to produce a short sentence to attract readers to read the news. One news article often contains multiple keyphrases that are of interest to different users, which can naturally have multiple reasonable headlines. However, most existing methods focus on the single headline generation. In this paper, we propose generating multiple headlines with keyphrases of user interests, whose main idea is to generate multiple keyphrases of interest to users for the news first, and then generate multiple keyphrase-relevant headlines. We propose a multi-source Transformer decoder, which takes three sources as inputs: (a) keyphrase, (b) keyphrase-filtered article, and (c) original article to generate keyphrase-relevant, high-quality, and diverse headlines. Furthermore, we propose a simple and effective method to mine the keyphrases of interest in the news article and build a first large-scale keyphrase-aware news headline corpus, which contains over 180K aligned triples of $<$news article, headline, keyphrase$>$. Extensive experimental comparisons on the real-world dataset show that the proposed method achieves state-of-the-art results in terms of quality and diversity

cs.CL↗

arXiv2020

Tell Me How to Ask Again: Question Data Augmentation with Controllable Rewriting in Continuous Space

In this paper, we propose a novel data augmentation method, referred to as Controllable Rewriting based Question Data Augmentation (CRQDA), for machine reading comprehension (MRC), question generation, and question-answering natural language inference tasks. We treat the question data augmentation task as a constrained question rewriting problem to generate context-relevant, high-quality, and diverse question data samples. CRQDA utilizes a Transformer autoencoder to map the original discrete question into a continuous embedding space. It then uses a pre-trained MRC model to revise the question representation iteratively with gradient-based optimization. Finally, the revised question representations are mapped back into the discrete space, which serve as additional question data. Comprehensive experiments on SQuAD 2.0, SQuAD 1.1 question generation, and QNLI tasks demonstrate the effectiveness of CRQDA

cs.CL↗

arXiv2020

Decoy Allocation Games on Graphs with Temporal Logic Objectives

We study a class of games, in which the adversary (attacker) is to satisfy a complex mission specified in linear temporal logic, and the defender is to prevent the adversary from achieving its goal. A deceptive defender can allocate decoys, in addition to defense actions, to create disinformation for the attacker. Thus, we focus on the problem of jointly synthesizing a decoy placement strategy and a deceptive defense strategy that maximally exploits the incomplete information the attacker about the decoy locations. We introduce a model of hypergames on graphs with temporal logic objectives to capture such adversarial interactions with asymmetric information. Using the hypergame model, we analyze the effectiveness of a given decoy placement, quantified by the set of deceptive winning states where the defender can prevent the attacker from satisfying the attack objective given its incomplete information about decoy locations. Then, we investigate how to place decoys to maximize the defender's deceptive winning region. Considering the large search space for all possible decoy allocation strategies, we incorporate the idea of compositional synthesis from formal methods and show that the objective function in the class of decoy allocation problem is monotone and non-decreasing. We derive the sufficient conditions under which the objective function for the decoy allocation problem is submodular, or supermodular, respectively. We show a sub-optimal allocation can be efficiently computed by iteratively composing the solutions of hypergames with a subset of decoys and the solution of a hypergame given a single decoy. We use a running example to illustrate the proposed method.

cs.GT↗

arXiv2020

A Theory of Hypergames on Graphs for Synthesizing Dynamic Cyber Defense with Deception

In this chapter, we present an approach using formal methods to synthesize reactive defense strategy in a cyber network, equipped with a set of decoy systems. We first generalize formal graphical security models--attack graphs--to incorporate defender's countermeasures in a game-theoretic model, called an attack-defend game on graph. This game captures the dynamic interactions between the defender and the attacker and their defense/attack objectives in formal logic. Then, we introduce a class of hypergames to model asymmetric information created by decoys in the attacker-defender interactions. Given qualitative security specifications in formal logic, we show that the solution concepts from hypergames and reactive synthesis in formal methods can be extended to synthesize effective dynamic defense strategy using cyber deception. The strategy takes the advantages of the misperception of the attacker to ensure security specification is satisfied, which may not be satisfiable when the information is symmetric.

cs.CR↗

arXiv2020

Dynamic Hypergames for Synthesis of Deceptive Strategies with Temporal Logic Objectives

In this paper, we study the use of deception for strategic planning in adversarial environments. We model the interaction between the agent (player 1) and the adversary (player 2) as a two-player concurrent game in which the adversary has incomplete information about the agent's task specification in temporal logic. During the online interaction, the adversary can infer the agent's intention from observations and adapt its strategy so as to prevent the agent from satisfying the task. To plan against such an adaptive opponent, the agent must leverage its knowledge about the adversary's incomplete information to influence the behavior of the opponent, and thereby being deceptive. To synthesize a deceptive strategy, we introduce a class of hypergame models that capture the interaction between the agent and its adversary given asymmetric, incomplete information. A hypergame is a hierarchy of games, perceived differently by the agent and its adversary. We develop the solution concept of this class of hypergames and show that the subjectively rationalizable strategy for the agent is deceptive and maximizes the probability of satisfying the task in temporal logic. This deceptive strategy is obtained by modeling the opponent evolving perception of the interaction and integrating the opponent model into proactive planning. Following the deceptive strategy, the agent chooses actions to influence the game history as well as to manipulate the adversary's perception so that it takes actions that benefit the goal of the agent. We demonstrate the correctness of our deceptive planning algorithm using robot motion planning examples with temporal logic objectives and design a detection mechanism to notify the agent of potential errors in modeling of the adversary's behavior.

cs.GT↗

arXiv2020

A Receding-Horizon MDP Approach for Performance Evaluation of Moving Target Defense in Networks

In this paper, we study the problem of assessing the effectiveness of a proactive defense-by-detection policy with a network-based moving target defense. We model the network system using a probabilistic attack graph--a graphical security model. Given a network system with a proactive defense strategy, an intelligent attacker needs to perform reconnaissance repeatedly to learn about the locations of intrusion detection systems and re-plan optimally to reach the target while avoiding detection. To compute the attacker's strategy for security evaluation, we develop a receding-horizon planning algorithm using a risk-sensitive Markov decision process with a time-varying reward function. Finally, we implement both defense and attack strategies in a synthetic network and analyze how the frequency of network randomization and the number of detection systems can influence the success rate of the attacker. This study provides insights for designing proactive defense strategies against online and multi-stage attacks by a resourceful attacker.

cs.CR↗

arXiv2020

Deceptive Labeling: Hypergames on Graphs for Stealthy Deception

With the increasing sophistication of attacks on cyber-physical systems, deception has emerged as an effective tool to improve system security and safety by obfuscating the attacker's perception. In this paper, we present a solution to the deceptive game in which a control agent is to satisfy a Boolean objective specified by a co-safe temporal logic formula in the presence of an adversary. The agent intentionally introduces asymmetric information to create payoff misperception, which manifests as the misperception of the labeling function in the game model. Thus, the adversary is unable to accurately determine which logical formula is satisfied by a given outcome of the game. We introduce a model called hypergame on graph to capture the asymmetrical information with one-sided payoff misperception. Based on this model, we present the solution of such a hypergame and use the solution to synthesize stealthy deceptive strategies. Specifically, deceptive sure winning and deceptive almost-sure winning strategies are developed by reducing the hypergame to a two-player game and one-player stochastic game with reachability objectives. A running example is introduced to demonstrate the game model and the solution concept used for strategy synthesis.

cs.GT↗

arXiv2020

A novel 3D multi-path DenseNet for improving automatic segmentation of glioblastoma on pre-operative multi-modal MR images

Convolutional neural networks have achieved excellent results in automatic medical image segmentation. In this study, we proposed a novel 3D multi-path DenseNet for generating the accurate glioblastoma (GBM) tumor contour from four multi-modal pre-operative MR images. We hypothesized that the multi-path architecture could achieve more accurate segmentation than a single-path architecture. 258 GBM patients were included in this study. Each patient had four MR images (T1-weighted, contrast-enhanced T1-weighted, T2-weighted, and FLAIR) and the manually segmented tumor contour. We built a 3D multi-path DenseNet that could be trained to generate the corresponding GBM tumor contour from the four MR images. A 3D single-path DenseNet was also built for comparison. Both DenseNets were based on the encoder-decoder architecture. All four images were concatenated and fed into a single encoder path in the single-path DenseNet, while each input image had its own encoder path in the multi-path DenseNet. The patient cohort was randomly split into a training set of 180 patients, a validation set of 39 patients, and a testing set of 39 patients. Model performance was evaluated using the Dice similarity coefficient (DSC), average surface distance (ASD), and 95% Hausdorff distance (HD95%). Wilcoxon signed-rank tests were conducted to examine the model differences. The single-path DenseNet achieved a DSC of 0.911$\pm$0.060, ASD of 1.3$\pm$0.7 mm, and HD95% of 5.2$\pm$7.1 mm, while the multi-path DenseNet achieved a DSC of 0.922$\pm$0.041, ASD of 1.1$\pm$0.5 mm, and HD95% of 3.9$\pm$3.3 mm. The p-values of all Wilcoxon signed-rank tests were less than 0.05. Both 3D DenseNets generated GBM tumor contours in good agreement with the manually segmented contours from multi-modal MR images. The multi-path DenseNet achieved more accurate tumor segmentation than the single-path DenseNet.

physics.med-ph↗

arXiv2020

RikiNet: Reading Wikipedia Pages for Natural Question Answering

Reading long documents to answer open-domain questions remains challenging in natural language understanding. In this paper, we introduce a new model, called RikiNet, which reads Wikipedia pages for natural question answering. RikiNet contains a dynamic paragraph dual-attention reader and a multi-level cascaded answer predictor. The reader dynamically represents the document and question by utilizing a set of complementary attention mechanisms. The representations are then fed into the predictor to obtain the span of the short answer, the paragraph of the long answer, and the answer type in a cascaded manner. On the Natural Questions (NQ) dataset, a single RikiNet achieves 74.3 F1 and 57.9 F1 on long-answer and short-answer tasks. To our best knowledge, it is the first single model that outperforms the single human performance. Furthermore, an ensemble RikiNet obtains 76.1 F1 and 61.3 F1 on long-answer and short-answer tasks, achieving the best performance on the official NQ leaderboard

cs.CL↗

arXiv2020

Role-Wise Data Augmentation for Knowledge Distillation

Knowledge Distillation (KD) is a common method for transferring the ``knowledge'' learned by one machine learning model (the \textit{teacher}) into another model (the \textit{student}), where typically, the teacher has a greater capacity (e.g., more parameters or higher bit-widths). To our knowledge, existing methods overlook the fact that although the student absorbs extra knowledge from the teacher, both models share the same input data -- and this data is the only medium by which the teacher's knowledge can be demonstrated. Due to the difference in model capacities, the student may not benefit fully from the same data points on which the teacher is trained. On the other hand, a human teacher may demonstrate a piece of knowledge with individualized examples adapted to a particular student, for instance, in terms of her cultural background and interests. Inspired by this behavior, we design data augmentation agents with distinct roles to facilitate knowledge distillation. Our data augmentation agents generate distinct training data for the teacher and student, respectively. We find empirically that specially tailored data points enable the teacher's knowledge to be demonstrated more effectively to the student. We compare our approach with existing KD methods on training popular neural architectures and demonstrate that role-wise data augmentation improves the effectiveness of KD over strong prior approaches. The code for reproducing our results can be found at https://github.com/bigaidream-projects/role-kd

cs.LG↗

arXiv2020

Interactive Machine Comprehension with Information Seeking Agents

Existing machine reading comprehension (MRC) models do not scale effectively to real-world applications like web-level information retrieval and question answering (QA). We argue that this stems from the nature of MRC datasets: most of these are static environments wherein the supporting documents and all necessary information are fully observed. In this paper, we propose a simple method that reframes existing MRC datasets as interactive, partially observable environments. Specifically, we "occlude" the majority of a document's text and add context-sensitive commands that reveal "glimpses" of the hidden text to a model. We repurpose SQuAD and NewsQA as an initial case study, and then show how the interactive corpora can be used to train a model that seeks relevant information through sequential decision making. We believe that this setting can contribute in scaling models to web-level QA scenarios.

cs.CL↗

arXiv2020

Feature Lenses: Plug-and-play Neural Modules for Transformation-Invariant Visual Representations

Convolutional Neural Networks (CNNs) are known to be brittle under various image transformations, including rotations, scalings, and changes of lighting conditions. We observe that the features of a transformed image are drastically different from the ones of the original image. To make CNNs more invariant to transformations, we propose "Feature Lenses", a set of ad-hoc modules that can be easily plugged into a trained model (referred to as the "host model"). Each individual lens reconstructs the original features given the features of a transformed image under a particular transformation. These lenses jointly counteract feature distortions caused by various transformations, thus making the host model more robust without retraining. By only updating lenses, the host model is freed from iterative updating when facing new transformations absent in the training data; as feature semantics are preserved, downstream applications, such as classifiers and detectors, automatically gain robustness without retraining. Lenses are trained in a self-supervised fashion with no annotations, by minimizing a novel "Top-K Activation Contrast Loss" between lens-transformed features and original features. Evaluated on ImageNet, MNIST-rot, and CIFAR-10, Feature Lenses show clear advantages over baseline methods.

cs.CV↗

arXiv2020

Learning to Locomote with Deep Neural-Network and CPG-based Control in a Soft Snake Robot

In this paper, we present a new locomotion control method for soft robot snakes. Inspired by biological snakes, our control architecture is composed of two key modules: A deep reinforcement learning (RL) module for achieving adaptive goal-tracking behaviors with changing goals, and a central pattern generator (CPG) system with Matsuoka oscillators for generating stable and diverse locomotion patterns. The two modules are interconnected into a closed-loop system: The RL module, analogizing the locomotion region located in the midbrain of vertebrate animals, regulates the input to the CPG system given state feedback from the robot. The output of the CPG system is then translated into pressure inputs to pneumatic actuators of the soft snake robot. Based on the fact that the oscillation frequency and wave amplitude of the Matsuoka oscillator can be independently controlled under different time scales, we further adapt the option-critic framework to improve the learning performance measured by optimality and data efficiency. The performance of the proposed controller is experimentally validated with both simulated and real soft snake robots.

cs.RO↗

arXiv2020

Secure-by-synthesis network with active deception and temporal logic specifications

This paper is concerned with the synthesis of strategies in network systems with active cyber deception. Active deception in a network employs decoy systems and other defenses to conduct defensive planning against the intrusion of malicious attackers who have been confirmed by sensing systems. In this setting, the defender's objective is to ensure the satisfaction of security properties specified in temporal logic formulas. We formulate the problem of deceptive planning with decoy systems and other defenses as a two-player games with asymmetrical information and Boolean payoffs in temporal logic. We use level-2 hypergame with temporal logic objectives to capture the incomplete/incorrect knowledge of the attacker about the network system as a payoff misperception. The true payoff function is private information of the defender. Then, we extend the solution concepts of $omega$-regular games to analyze the attacker's rational strategy given her incomplete information. By generalizing the solution of level-2 hypergame in the normal form to extensive form, we extend the solutions of games with safe temporal logic objectives to decide whether the defender can ensure security properties to be satisfied with probability one, given any possible strategy that is perceived to be rational by the attacker. Further, we use the solution of games with co-safe (reachability) temporal logic objectives to determine whether the defender can engage the attacker, by directing the attacker to a high-fidelity honeypot. The effectiveness of the proposed synthesis methods is illustrated with synthetic network systems with honeypots.

cs.GT↗

arXiv2020

Synthesis of Deceptive Strategies in Reachability Games with Action Misperception

We consider a class of two-player turn-based zero-sum games on graphs with reachability objectives, known as reachability games, where the objective of Player 1 (P1) is to reach a set of goal states, and that of Player 2 (P2) is to prevent this. In particular, we consider the case where the players have asymmetric information about each other's action capabilities: P2 starts with an incomplete information (misperception) about P1's action set, and updates the misperception when P1 uses an action previously unknown to P2. When P1 is made aware of P2's misperception, the key question is whether P1 can control P2's perception so as to deceive P2 into selecting actions to P1's advantage? We show that there might exist a deceptive winning strategy for P1 that ensures P1's objective is achieved with probability one from a state otherwise losing for P1, had the information being symmetric and complete. We present three key results: First, we introduce a dynamic hypergame model to capture the reachability game with evolving misperception of P2. Second, we present a fixed-point algorithm to compute the Deceptive Almost-Sure Winning (DASW) region and DASW strategy. Finally, we show that DASW strategy is at least as powerful as Almost-Sure Winning (ASW) strategy in the game in which P1 does not account for P2's misperception. We illustrate our algorithm using a robot motion planning in an adversarial environment.

cs.GT↗