热点资讯
app开发市场价格 信得过东说念主工智能
发布日期：2024-07-18 14:28 点击次数：193
app
福彩快乐8第2024175期(上周三)开奖回顾：07 09 12 15 17 19 32 33 40 47 48 49 55 58 62 65 66 69 70 73，其中奖号遗漏总值为60，冷温热码比为3：6：11。
整场比赛，中国队一攻失误17次，而马刺队只有7次。其中杨瀚森和程帅澎分别失误4次，廖三宁和贺希宁各有3次失误。
信得过东说念主工智能NASAC/FMAC
跟着东说念主类活命被东说念主工智能芜俚浸透，公众禁受东说念主工智能的进程越来越高，东说念主工智能的信得过初始受到芜俚的存眷。东说念主工智能的“信得过性”是在可靠性、安全性、鲁棒性、可讲授性、可控性等宽绰想法的基础上发展起来的一个新想法。在智能系统中，“信得过”是指一个智能实体在盛开、动态环境下已毕目的需求的动态历程中，其行径及产生收尾老是相宜东说念主们的预期。它强调目的与已毕相符，强调行径和收尾的可讲授性、可展望性和可适度性。除此以外，信得过智能计较还条款在受到诸如环境影响、外部膺惩等动态环境骚动时，仍然粗略握续提供相宜预期的处事。关于东说念主工智能的信得过问题，现在国表里揣摸尚处于摸索阶段。本次论坛将从表面、门径、时刻、诓骗等多个维度先容现在最新的信得逾期刻，旨在为提高智能系统建立服从、改善软硬件假想质料、增强智能系统信得过、优化建立历程和适度资本方面提供念念路。论坛将邀请学术界群众先容高水平揣摸收尾，并开展联系的时刻念念辨盘考。
论坛组织委员会：
李 钦（华东师范大学）
陈仪香（华东师范大学）
日程安排：
时 间：2020年11月21日（星期六）14:00-17:30
地 点：重庆富力沐日货仓 会议室1
论坛议程：
弘扬及嘉宾简介：
1. 张立军：Deep Neural Networks: From Verification to Generalisation
纲目：In this talk, we survey major techniques and tools for verifying deep neural networks, in particular, the notion of robustness for specifying safety properties. Moreover, we study the novel concept of weight correlation in deep neural networks and discusses its impact on the networks’ generalisation ability. We argue that weight correlation can improve generalisation of neural networks. Finally, we discuss how it can be intertwined with robustness to get reliable networks.
简介：Lijun Zhang is a research professor at State Key Laboratory of Computer Science, Institute of Software Chinese Academy of Sciences (ISCAS). He is also the director of the Sino-Europe Joint Institute of Dependable and Smart Software at the Institute of Intelligent Software in Guangzhou. Before this he was a postdoctoral researcher at University of Oxford, and then an associate professor at Language-Based Technology section, DTU Compute, Technical University of Denmark. He gained a Diploma Degree and a PhD (Dr. Ing.) at Saarland University. His research interests include: probabilistic model checking, simulation reduction and decision algorithms, abstraction and model checking, learning algorithms, and verification of deep neural networks. He is leading the development of several tools including PRODeep for verifying neural networks, ROLL for learning automata, and the model checker ePMC, previously known as IscasMC.
2. 卜磊：基于无梯度优化的神经收集叛逆样本生成
纲目：连年来以图像识别为代表的东说念主工智能系统芜俚诓骗在各个边界。然则，相关连统与模子由于内在不能讲授性，其行径极易收到数据骚动。若何对相关连统进行叛逆样本生成，对意会联系模子行径，隐没风险具有伏击好奇好奇。咱们面向黑盒图像识别模子的叛逆样本生成进行了探索与翻新，建议一种基于无梯度优化的黑盒模子叛逆样本生成门径，将叛逆样本生成转机为优化问题，并基于智能化门径进行求解，基于此念念路咱们假想与已毕了联系器具DFA，并在公认案例集上佐证其科罚才调。
简介：卜磊 南京大学计较机科学与时刻系耕种，博士生导师。主要揣摸边界是软件工程与体式化门径，包括模子磨练时刻，及时混成系统，信息物理交融系统等标的。2010年在南京大学获取计较机博士学位。曾在CMU、MSRA、UTD、FBK等科研机构进行访学与配合揣摸。现在已在联系边界伏击期刊与会议如TCAD、TC、TCPS、TPDS、RTSS、CAV等上发表论文五十余篇。入选中国计较机学会后生东说念主才发展蓄意，微软亚洲揣摸院铸星蓄意，app开发软件价格NASAC后生软件翻新奖等。
3. 张民：Accelerating Robustness Verification of Deep Neural Networks with Lazy&Eager Falsification
纲目: Bad scalability is one of the challenging problems to the robustness verification of neural networks. In this talk, we will introduce a simple yet effective approach to accelerate the robustness verification by lazy\&eager falsification. In the approach, we divide the robustness verification problem into sub-problems with respect to target labels. Once the property is falsified for a specific label, we can safely conclude that the neural network is non-robust. The falsification is both lazy and eager. Being lazy means that a sub-problem is not falsified unless it has to and being eager means that the sub-problems that are more likely to be falsifiable have higher priority to falsify. We leverage symbolic interval propagation and linear relaxation techniques to determine the order of the sub-problems for falsification. Our approach is orthogonal to, and can be integrated with existing verification techniques. We integrate it with four state-of-the-art verification tools, i.e., MipVerify, Neurify, DeepZ, and DeepPoly. Experimental results show that our approach can significantly improve these tools, up to 200x speedup, when the perturbation distance is set in a reasonable range.
简介：张民于2011年在日本北陆先端科学时刻大学院大学赢得博士学位，2011至2014年在JAIST软件考据中心从事博士后揣摸职责，2014年起加入华东师范大学软件工程学院。主要揣摸边界包括体式化门径与信得过计较表面，将体式化门径诓骗于智能系统，物联网，镶嵌式系统的信得过考据与分析，联系职责发表在包括ETAPS、DAC、DATE、TCAD、软件学报等会议和期刊上，赢得APSEC2019唯独最好论文奖，DCIT2015最好论文奖。曾担任TASE2017、FM4AI2019法子委员会主席，DATE2021，ICFEM2016等海外盛名学术会议法子委员，SCP客座裁剪。主握国度当然科学基金面上姿色，后生姿色，中法海外配合“蔡元培”姿色等姿色。CCF体式化专委会委员。
4. 刘万伟：Verifying ReLU Neural Networks from a Model Checking Perspective
纲目：Neural networks, as an important computing model, have a wide application in artificial intelligence (AI) domain. From the perspective of computer science, such a computing model requires a formal description of its behaviors, particularly the relation between input and output. In addition, such specifications ought to be verified automatically. ReLU (rectified linear unit) neural networks are intensively used in practice. In this paper, we present ReLUTemporal Logic (ReTL), whose semantics is defined with respect to ReLUneural networks, which could specify valuerelated properties about the network. We show that the model checking algorithm for the Σ2∪Π2 fragment of ReTL, which can express properties such as output reachability, is decidable in EXPSPACE. We have also implemented our algorithm with a prototype tool, and experimental results demonstrate the feasibility of the presented model checking approach.
简介：Wanwei Liu received his Ph.D degree in National University of Defense Technology in 2009, he is an associated professor in College of Computer Science, National University of Defense Technology. He is a senior member of CCF. His research interests include theoretical computer science (particularly in automata theory and temporal logic), formal methods (particularly in verification), and software engineering. His work has been published on TSE, ICSE, ASE, TACAS, IJCAI. His work acquires Gold prize (1st prize) in TACAS SV-Comp verification tool track multiple times.
5. 薛白：PAC Model Checking of Block-Box Continuous-Time Dynamical Systems
纲目：In this talk I will present a model checking approach to finite-time safety verification of black-box continuous-time dynamical systems within the framework of probably approximately correct (PAC) learning. The black-box dynamical systems are the ones, for which no model is given but whose states changing continuously through time within a finite time interval can be observed at some discrete time instants for a given input. The new model checking approach is termed as PAC model checking due to incorporation of learned models with correctness guarantees expressed using the terms error probability and confidence. Based on the error probability and confidence level, our approach provides statistically formal guarantees that the time-evolving trajectories of the black-box dynamical system over finite time horizons fall within the range of the learned model plus a bounded interval, contributing to insights on the reachability of the black-box system and thus on the satisfiability of its safety requirements. The learned model together with the bounded interval is obtained by scenario optimization, which boils down to a linear programming problem.
简介：Dr. Bai Xue is an associate research professor at State Key Laboratory of Computer Science, Institute of Software Chinese Academy of Sciences since November, 2017. He received the B.Sc. degree in information and computing science from Tianjin University of Technology and Education in 2008, and the Ph.D. degree in applied mathematics from Beihang University in 2014. Prior to joining Institute of Software, he worked as a research fellow in the Centre for High Performance Embedded Systems at Nanyang Technological University from May, 2014 to September, 2015, and as a postdoc in the Department füer Informatik at Carl von Ossietzky Universität Oldenburg from November, 2015 to October, 2017.
6. 孟国柱：深度神经收集的安全遏制与膺惩推行
纲目：连年来以深度学习为代表的东说念主工智能时刻得到飞速发展和践诺，但神经收集存在的安全弱势给社会和个东说念主带来很大的风险和遏制。为了笃定深度神经收集的弱势，咱们开展了一项针对神经收集安全遏制的详尽性调研，从模子萃取、模子逆向、模子投毒和叛逆膺惩四个方面谈判常见的膺惩时刻以及在多个量化目的中的对比。在实证分析中发现，由于攻防两头的信息别离称，在黑盒膺惩中膺惩者一样需要构造大宗样本并与目的模子进行大宗交互查询，制约着黑盒膺惩的服从。因此咱们摄取了四种数据约减时刻来镌汰数据的冗余和进步黑盒膺惩服从。通过在多个数据集和复杂收集上进行测试，取得比SOTA更好的效果。终末凭证据验数据，进一步谈判了模子西宾的可讲授性。
简介：孟国柱，2017年博士毕业于新加坡南洋理工大学。于2018年9月加入中国科学院信息工程揣摸所任副揣摸员。曾获2019年ACM SIGSAC中国科技新星，赢得过2018年CCF-A类会议ICSE最好论文奖；联系揣摸职责也曾在软件工程和信息安全边界的海外顶级会议和期刊如ICSE，FSE，ASE，ISSTA等发表越过30篇著述。揣摸边界包括东说念主工智能系统安全与秘密，软件舛讹分析和检测，移动诓骗安全等。