基于对抗样本的人工智能反制技术研究进展与趋势

随着基于深度学习的人工智能技术的快速发展及其广泛应用,人们对其安全性的关注也日益凸显。特别是,最近一系列研究表明基于深度学习模型的人工智能系统容易受到对抗样本的攻击。对抗样本通过向正常样本中添加精心设计、人类难以察觉的微小扰动,可导致深度学习模型的严重误判。本文回顾基于对抗性图像和音频两类人工智能反制技术最新进展,并对这些研究成果进行分类和综合比较,最后对现有挑战与未来研究趋势进行了讨论和展望。     

Knowledge acquisition techniques for feature recognition in CAD models

Automatic Feature Recognition (AFR) techniques are an important tool for achieving a true integration of design and manufacturing stages during the product development. In particular, AFR systems offer capabilities for recognising high-level geometrical entities, features, in Computer-Aided Design (CAD) models. However, the recognition performances of most of the existing AFR systems are limited to the requirements of specific applications. This paper presents automatic knowledge acquisition techniques to support the development of AFR systems that could be deployed in different application domains. In particular, a method to generate automatically feature recognition rules is proposed. These rules are formed by applying an inductive learning algorithm on training data consisting of feature examples. In addition, a technique for defining automatically feature hints from such rule sets is described. The knowledge acquisition techniques presented in this study are implemented within a prototype feature recognition system and its capabilities are verified on two benchmarking parts.     

基于深度学习的中文命名实体识别最新研究进展综述

中文命名实体识别(CNER)任务是问答系统、机器翻译、信息抽取等自然语言应用的基础底层任务。传统的CNER系统借助人工设计的领域词典和语法规则,取得了不错的实验效果,但存在泛化能力弱、鲁棒性差、维护难等缺点。近年来兴起的深度学习技术通过端到端的方式自动提取文本特征,弥补了上述不足。该文对基于深度学习的中文命名实体识别任务最新研究进展进行了综述,先介绍中文命名实体识别任务的概念、应用现状和难点,接着简要介绍中文命名实体识别任务的常用数据集和评估方法,并按照主要网络架构对中文命名实体识别任务上的深度学习模型进行分类和梳理,最后对这一任务的未来研究方向进行了展望。     

Learning domain ontologies for semantic Web service descriptions

High quality domain ontologies are essential for successful employment of semantic Web services. However, their acquisition is difficult and costly, thus hampering the development of this field. In this paper we report on the first stage of research that aims to develop (semi-)automatic ontology learning tools in the context of Web services that can support domain experts in the ontology building task. The goal of this first stage was to get a better understanding of the problem at hand and to determine which techniques might be feasible to use. To this end, we developed a framework for (semi-)automatic ontology learning from textual sources attached to Web services. The framework exploits the fact that these sources are expressed in a specific sublanguage, making them amenable to automatic analysis. We implement two methods in this framework, which differ in the complexity of the employed linguistic analysis. We evaluate the methods in two different domains, verifying the quality of the extracted ontologies against high quality hand-built ontologies of these domains.

Our evaluation lead to a set of valuable conclusions on which further work can be based. First, it appears that our method, while tailored for the Web services context, might be applicable across different domains. Second, we concluded that deeper linguistic analysis is likely to lead to better results. Finally, the evaluation metrics indicate that good results can be achieved using only relatively simple, off the shelf techniques. Indeed, the novelty of our work is not in the used natural language processing methods but rather in the way they are put together in a generic framework specialized for the context of Web services.     

An application of learning to robotic planning

Conventional approaches to robotic planning have focused on the resolution theorem prover, using general-purpose search heuristics, with the desired goal expressed in terms of logical calculus. These approaches suffer from several drawbacks; one major problem encountered in these approaches is the speed of planning. In this paper we describe an approach of applying supervised learning to robotic planning. The learning system is an intermediate one between rote learning and generalization learning, and is based on the concept of analogy. Simulation examples of various robot tasks are presented to demonstrate the significant increase in the systems's planning speed and to compare it with some existing systems.This work was supported by the National Science Foundation grant ENG-74-17586.     

Explanation-Based Generalization: A Unifying View

The problem of formulating general concepts from specific training examples has long been a major focus of machine learning research. While most previous research has focused on empirical methods for generalizing from a large number of training examples using no domain-specific knowledge, in the past few years new methods have been developed for applying domain-specific knowledge to for-mulate valid generalizations from single training examples. The characteristic common to these methods is that their ability to generalize from a single example follows from their ability to explain why the training example is a member of the concept being learned. This paper proposes a general, domain-independent mechanism, called EBG, that unifies previous approaches to explanation-based generalization. The EBG method is illustrated in the context of several example problems, and used to contrast several existing systems for explanation-based generalization. The perspective on explanation-based generalization afforded by this general method is also used to identify open research problems in this area.     

Applying Learning by Examples for Digital Design Automation

This paper describes a new learning by example mechanism and its application for digital circuit design automation. This mechanism uses finite state machines to represent the inferred models or designs. The resultant models are easy to be implemented in hardware using current VLSI technologies. Our simulation results show that it is often possible to infer a well-defined deterministic model or design from just one sequence of examples. In addition this mechanism is able to handle sequential task involving long-term dependence. This new learning by example mechanism is used as a design by example system for automatic synthesis of digital circuits. Such systems have not previously been successfully developed mainly because of the lack of mechanism to implement them. From artificial neural network research, it seems possible to apply the knowledge gained from learning by example to form a design by example system. However, one of the problems with neural network approaches is that the resultant models are very difficult to be implemented in hardware using current VLSI technologies. By using the mechanism described in this paper, the resultant models are finite state machines that are well suited for digital designs. Several sequential circuit design examples are simulated and tested. Although our test results show that such a system is feasible for designing simple circuits or small-scale circuit modules, the feasibility of such a system for large-scale circuit design remains to be showed. Both the learning mechanism and the design method show potential and the future research directions are provided.     

元学习研究综述

深度学习在大量领域取得优异成果,但仍然存在着鲁棒性和泛化性较差、难以学习和适应未观测任务、极其依赖大规模数据等问题.近两年元学习在深度学习上的发展,为解决上述问题提供了新的视野.元学习是一种模仿生物利用先前已有的知识,从而快速学习新的未见事物能力的一种学习定式.元学习的目标是利用已学习的信息,快速适应未学习的新任务.这...     

自动语音辨识对抗攻击和防御技术综述

语音辨识技术是人机交互的重要方式。随着深度学习的不断发展,基于深度学习的自动语音辨识系统也取得了重要进展。然而,经过精心设计的音频对抗样本可以使得基于神经网络的自动语音辨识系统产生错误,给基于语音辨识系统的应用带来安全风险。为了提升基于神经网络的自动语音辨识系统的安全性,需要对音频对抗样本的攻击和防御进行研究。基于此,分析总结对抗样本生成和防御技术的研究现状,介绍自动语音辨识系统对抗样本攻击和防御技术面临的挑战和解决思路。     

A New Model of Computation for Learning Vision Modules from Examples

This paper addresses an important class of mimicry problems, where the goal is to construct a computer program which is functionally equivalent to an observed behaviour. Computer vision research can be considered such a challenge, where a researcher attempts to impart human visual abilities to a computer. Unfortunately this has proved a difficult task, not least because our vision processes occur mostly at a subconscious level. It is therefore useful to study the general mimicry problem in order to develop tools which may assist computer vision research.This paper formalises a mimicry problem as one in which a computer learning system (L) constructs a solution from a given program structure (i.e. template or outline) by posing questions to an Oracle. The latter is an entity which, when given an input value, produces the corresponding output of the function which is to be mimicked.In order to define a program's structure, particularly one which can be extracted from any computer program automatically, a new model of computation is developed. Based on this a fast algorithm which determines the best questions to pose to the Oracle is then described. Thus L relieves the human programmer of the difficulties faced in choosing the examples from which to learn. This is important because a human programmer might inadvertently choose biased, redundant or otherwise unhelpful examples. Results are shown which demonstrate the utility of a complete learning system (L) based on this work.This paper represents background theory and initial algorithms which further work will extend into powerful automatic learning systems, examples of which are found in [36] and [38].     

Explicit representation of terms defined by counter examples

Anti-unification guarantees the existence of a term which is an explicit representation of the most specific generalization of a collection of terms. This provides a formal basis for learning from examples. Here we address the dual problem of computing a generalization given a set of counter examples. Unlike learning from examples an explicit, finite representation for the generalization does not always exist. We show that the problem is decidable by providing an algorithm which, given an implicit representation will return a finite explicit representation or report that none exists. Applications of this result to the problem of negation as failure and to the representation of solutions to systems of equations and inequations are also mentioned.Research performed while visiting from the: Dept. of Computer Science, University of Melbourne, Parkville 3052, Australia.     

Learning classifier systems: a survey

Learning classifier systems (LCSs) are rule- based systems that automatically build their ruleset. At the origin of Holland’s work, LCSs were seen as a model of the emergence of cognitive abilities thanks to adaptive mechanisms, particularly evolutionary processes. After a renewal of the field more focused on learning, LCSs are now considered as sequential decision problem-solving systems endowed with a generalization property. Indeed, from a Reinforcement Learning point of view, LCSs can be seen as learning systems building a compact representation of their problem thanks to generalization. More recently, LCSs have proved efficient at solving automatic classification tasks. The aim of the present contribution is to describe the state-of- the-art of LCSs, emphasizing recent developments, and focusing more on the sequential decision domain than on automatic classification.     

基于采样策略的主动学习算法研究进展

主动学习算法通过选择信息含量大的未标记样例交由专家进行标记,多次循环使分类器的正确率逐步提高,进而在标记总代价最小的情况下获得分类器的强泛化能力,这一技术引起了国内外研究人员的关注.侧重从采样策略的角度,详细介绍了主动学习中学习引擎和采样引擎的工作过程,总结了主动学习算法的理论研究成果,详细评述了主动学习的研究现状和发展动态.首先,针对采样策略选择样例的不同方式将主动学习算法划分为不同类型,进而,对基于不同采样策略的主动学习算法进行了深入地分析和比较,讨论了各种算法适用的应用领域及其优缺点.最后指出了存在的开放性问题和进一步的研究方向.     

Exploratory learning structures in artificial cognitive systems

The major goal of the COSPAL project is to develop an artificial cognitive system architecture, with the ability to autonomously extend its capabilities. Exploratory learning is one strategy that allows an extension of competences as provided by the environment of the system. Whereas classical learning methods aim at best for a parametric generalization, i.e., concluding from a number of samples of a problem class to the problem class itself, exploration aims at applying acquired competences to a new problem class, and to apply generalization on a conceptual level, resulting in new models. Incremental or online learning is a crucial requirement to perform exploratory learning.In the COSPAL project, we mainly investigate reinforcement-type learning methods for exploratory learning, and in this paper we focus on the organization of cognitive systems for efficient operation. Learning is used over the entire system. It is organized in the form of four nested loops, where the outermost loop reflects the user-reinforcement-feedback loop, the intermediate two loops switch between different solution modes at symbolic respectively sub-symbolic level, and the innermost loop performs the acquired competences in terms of perception–action cycles. We present a system diagram which explains this process in more detail.We discuss the learning strategy in terms of learning scenarios provided by the user. This interaction between user (’teacher’) and system is a major difference to classical robotics systems, where the system designer places his world model into the system. We believe that this is the key to extendable robust system behavior and successful interaction of humans and artificial cognitive systems.We furthermore address the issue of bootstrapping the system, and, in particular, the visual recognition module. We give some more in-depth details about our recognition method and how feedback from higher levels is implemented. The described system is however work in progress and no final results are available yet. The available preliminary results that we have achieved so far, clearly point towards a successful proof of the architecture concept.     

The State of the Art in Enhancing Trust in Machine Learning Models with the Use of Visualizations

Machine learning (ML) models are nowadays used in complex applications in various domains, such as medicine, bioinformatics, and other sciences. Due to their black box nature, however, it may sometimes be hard to understand and trust the results they provide. This has increased the demand for reliable visualization tools related to enhancing trust in ML models, which has become a prominent topic of research in the visualization community over the past decades. To provide an overview and present the frontiers of current research on the topic, we present a State-of-the-Art Report (STAR) on enhancing trust in ML models with the use of interactive visualization. We define and describe the background of the topic, introduce a categorization for visualization techniques that aim to accomplish this goal, and discuss insights and opportunities for future research directions. Among our contributions is a categorization of trust against different facets of interactive ML, expanded and improved from previous research. Our results are investigated from different analytical perspectives: (a) providing a statistical overview, (b) summarizing key findings, (c) performing topic analyses, and (d) exploring the data sets used in the individual papers, all with the support of an interactive web-based survey browser. We intend this survey to be beneficial for visualization researchers whose interests involve making ML models more trustworthy, as well as researchers and practitioners from other disciplines in their search for effective visualization techniques suitable for solving their tasks with confidence and conveying meaning to their data.     

零样本学习综述

尽管自深度学习发展以来,减少大量人工标记样本的需求使得零样本学习取得了不错的进展,以至于已经拥有比较完善的理论体系。但是对于零样本学习应用的研究寥寥无几,如何有效地对应用领域进行梳理是现阶段急需解决的问题。对零样本的理论体系进行介绍,通过一个例子引出零样本学习的定义,继而与广义零样本、监督学习比较,再而列举4个关键问题以及现有的解决方案,给出文本、图像、视频三方面常用的数据集;按照关键技术（属性、嵌入以及生成模型）出现时间顺序,对13个典型模型如何进行零样本学习展开描述,并对优点、缺点、创新点、挑选数据集以及表现进行总结;从词、图像、视频3个维度详细介绍了零样本学习在各个领域的应用;提出了零样本学习过程中出现的挑战并给出了对应的潜在研究方向。     

智能交互的物体识别增量学习技术综述

智能交互系统是研究人与计算机之间进行交流与通信,使计算机能够在最大程度上完成交互者的某个指令的一个领域。其发展的目标是实现人机交互的自主性、安全性和友好性。增量学习是实现这个发展目标的一个途径。本文对智能交互系统的任务、背景和获取信息来源进行简要介绍,主要对增量学习领域的已有工作进行综述。增量学习是指一个学习系统能不断地从新样本中学习新的知识,非常类似于人类自身的学习模式。它使智能交互系统拥有自我学习,提高交互体验的能力。文中对主要的增量学习算法的基本原理和特点进行了阐述,分析各自的优点和不足,并对进一步的研究方向进行展望。     

Cooperative database systems

In recent years, there has been considerable interest in techniques that enhance the cooperative behavior of database systems and many different techniques have been developed. These techniques approach the goal of cooperation in diverse ways; the differences may be in the specific task in which they offer cooperation, in the details of the solution, and even in the very interpretation of cooperative behavior. In this article we classify many different techniques into categories of cooperation, and we survey the techniques in some of these categories. Finally, we consider the challenges that remain and offer directions for new research. © 1996 John Wiley & Sons, Inc.     

经典线性算法的非线性核形式

经典线性算法的非线性核形式是近10年发展起来的一类非线性机器学习技术．它们最显著的特点是利用满足Mercer条件的核函数巧妙地推导出线性算法的非线性形式。并表述为与样本数目有关、与维数无关的优化问题．为了提高数值计算的稳定性、控制算法的推广能力以及改善迭代过程的收敛性。部分算法还采用了正则化技术．在概述核思想与核函数、正则化技术的基础上,系统地介绍了经典线性算法的非线性核形式,同时分析它们的优缺点,井讨论了进一步发展的方向．     

Improving Generalization with Active Learning

Active learning differs from learning from examples in that the learning algorithm assumes at least some control over what part of the input domain it receives information about. In some situations, active learning is provably more powerful than learning from examples alone, giving better generalization for a fixed number of training examples.In this article, we consider the problem of learning a binary concept in the absence of noise. We describe a formalism for active concept learning calledselective sampling and show how it may be approximately implemented by a neural network. In selective sampling, a learner receives distribution information from the environment and queries an oracle on parts of the domain it considers useful. We test our implementation, called anSG-network, on three domains and observe significant improvement in generalization.A preliminary version of this article appears as Cohn et al. (1990).