Neural and Evolutionary ComputingClassification Accuracy Improvement for Neuromorphic Computing Systems with One-level Precision SynapsesYandan Wang,Wei Wen,Linghao Song,Hai LiComments: Best Paper Award of ASP-DAC 2017Subjects:Neural and Evolutionary Computing (cs.NE)Brain inspired neuromorphic computing has demonstrated remarkable advantagesover traditional von Neumann architecture for its high energy efficiency andparallel data processing. However, the limited resolution of synaptic weightsdegrades system accuracy and thus impedes the use of neuromorphic systems. Inthis work, we propose thr ...继续阅读 (20)

Neural and Evolutionary ComputingMembrane-Dependent Neuromorphic Learning Rule for Unsupervised Spike Pattern DetectionSadique Sheik,Somnath Paul,Charles Augustine,Gert CauwenberghsComments: Published in IEEE BioCAS 2016 Proceedings, BioMedical Circuits and Systems Conference, 2016Subjects:Neural and Evolutionary Computing (cs.NE)Several learning rules for synaptic plasticity, that depend on either spiketiming or internal state variables, have been proposed in the past impartingvarying computational capabilities to Spiking Neural Networks. Due to designcomplications these learning rules are t ...继续阅读 (10)

Neural and Evolutionary ComputingGenerating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural NetworksMarwin H.S. Segler,Thierry Kogej,Christian Tyrchan,Mark P. WallerComments: 17 pages, 17 figuresSubjects:Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Learning (cs.LG); Chemical Physics (physics.chem-ph); Machine Learning (stat.ML)In de novo drug design, computational strategies are used to generate novelmolecules with good affinity to the desired biological target. In this work, weshow that recurrent neural networks can be trained as generative ...继续阅读 (2)

1. 新智元(3)MIT TR : 2017 可能是中国成为人工智能产业主力军的一年机器学习模型出错的四大原因及如何纠错【确认】Master就是AlphaGo升级版！60连胜背后看专家怎么评说2. 机器之心(3)深度 | 听！具备深度学习能力的助听器是怎么工作的？重磅 | DeepMind官方确认Master身份：全面回顾AlphaGo的再度出山之旅业界 | 认知计算和知识工作的革命 ...继续阅读 (3)

Neural and Evolutionary ComputingPlatEMO: A MATLAB Platform for Evolutionary Multi-Objective OptimizationYe Tian,Ran Cheng,Xingyi Zhang,Yaochu JinComments: 20 pages, 12 figures, 4 tablesSubjects:Neural and Evolutionary Computing (cs.NE)Over the last three decades, a large number of evolutionary algorithms havebeen developed for solving multiobjective optimization problems. However, therelacks an up-to-date and comprehensive software platform for researchers toproperly benchmark existing algorithms and for practitioners to apply selectedalgorithms to solve their real-world problems. The demand ...继续阅读 (7)

Neural and Evolutionary ComputingSimulated Tornado OptimizationS. Hossein Hosseini,Tohid Nouri,Afshin Ebrahimi,S. Ali HosseiniComments: 6 pages, 15 figures, 1 table, IEEE International Conference on Signal Processing and Intelligent System (ICSPIS16), Dec. 2016Subjects:Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)We propose a swarm-based optimization algorithm inspired by air currents of atornado. Two main air currents – spiral and updraft – are mimicked. Spiralmotion is designed for exploration of new search areas an ...继续阅读 (5)

Computer Vision and Pattern RecognitionRetrieving Similar X-Ray Images from Big Image Data Using Radon Barcodes with Single ProjectionsMorteza Babaie,H.R. Tizhoosh,Shujin Zhu,M.E. ShiriComments: Accepted for publication in ICPRAM 2017: The International Conference on Pattern Recognition Applications and Methods, Porto, Portugal, 2017Subjects:Computer Vision and Pattern Recognition (cs.CV)The idea of Radon barcodes (RBC) has been introduced recently. In this paper,we propose a content-based image retrieval approach for big datasets based onRadon barcodes. Our method (Single Projection Radon Ba ...继续阅读 (8)

Neural and Evolutionary ComputingAdult Content Recognition from Images Using a Mixture of Convolutional Neural NetworksMundher Al-Shabi,Tee Connie,Andrew Beng Jin TeohSubjects:Machine Learning (stat.ML); Computer Vision and Pattern Recognition (cs.CV); Neural and Evolutionary Computing (cs.NE)With rapid development of the Internet, the web contents become huge. Most ofthe websites are publicly available and anyone can access the contentseverywhere such as workplace, home and even schools. Nev-ertheless, not all theweb contents are appropriate for all users, especially children. An example oft ...继续阅读 (6)

官网：http://stanford.edu/class/cs229t/编者按：这是一个关于机器学习算法的理论分析课程，可以看作是线性代数、概率、优化的混合体。主要介绍asymptotics、uniform convergence、kernel methods和online learning四个部分。Instructor:Percy LiangNotes:https://web.stanford.edu/class/cs229t/notes.pdfSchedule(subject to change)Week 1Mon 1/4:Lecture 1: overview, asymptoticsWed 1/6:Lecture 2: asymptoticsFri 1/8:Section: probability/optimization reviewWeek 2Mon 1/11:Lecture 3: asymptoticsMon 1/11: Homework 0 (warmup)dueWed 1/13:Lecture 4: asymptoticsFri 1/15:Section: asymptoticsWeek 3Mon 1/18: no lecture (Martin Luther King Day)Wed 1/20:Lecture 5: uniform convergenceWed ...继续阅读 (6)

Neural and Evolutionary ComputingDeep neural heart rate variability analysisTamas MadlComments: 6 pages in NIPS 2016 Workshop on Machine Learning for Health (ML4HC)Subjects:Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Learning (cs.LG)Despite of the pain and limited accuracy of blood tests for early recognitionof cardiovascular disease, they dominate risk screening and triage. On theother hand, heart rate variability is non-invasive and cheap, but notconsidered accurate enough for clinical practice. Here, we tackle heart beatinterval based classification with deep ...继续阅读 (7)

Neural and Evolutionary ComputingImproving Human-Machine Cooperative Visual Search With Soft HighlightingRonald T. Kneusel,Michael C. MozerSubjects:Human-Computer Interaction (cs.HC); Neural and Evolutionary Computing (cs.NE)Advances in machine learning have produced systems that attain human-levelperformance on certain visual tasks, e.g., object identification. Nonetheless,other tasks requiring visual expertise are unlikely to be entrusted to machinesfor some time, e.g., satellite and medical imagery analysis. We describe ahuman-machine cooperative approach to visual search, the aim of which ...继续阅读 (11)

Information Retrieval"What is Relevant in a Text Document?": An Interpretable Machine Learning ApproachLeila Arras,Franziska Horn,Grégoire Montavon,Klaus-Robert Müller,Wojciech SamekComments: 19 pages, 7 figuresSubjects:Computation and Language (cs.CL); Information Retrieval (cs.IR); Learning (cs.LG); Machine Learning (stat.ML)Text documents can be described by a number of abstract concepts such assemantic category, writing style, or sentiment. Machine learning (ML) modelshave been trained to automatically map documents to these abstract concepts,allowing to annotate very large text ...继续阅读 (7)

Neural and Evolutionary ComputingHighway and Residual Networks learn Unrolled Iterative EstimationKlaus Greff,Rupesh K. Srivastava,Jürgen SchmidhuberComments: 10 + 3pages, under review for ICLR 2017Subjects:Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Learning (cs.LG)The past year saw the introduction of new architectures such as Highwaynetworks and Residual networks which, for the first time, enabled the trainingof feedforward networks with dozens to hundreds of layers using simple gradientdescent. While depth of representation has been posited as a primary ...继续阅读 (4)

Neural and Evolutionary ComputingScale-invariance of ruggedness measures in fractal fitness landscapesHendrik RichterSubjects:Chaotic Dynamics (nlin.CD); Neural and Evolutionary Computing (cs.NE); Populations and Evolution (q-bio.PE)The paper deals with using chaos to direct trajectories to targets andanalyzes ruggedness and fractality of the resulting fitness landscapes. Thetargeting problem is formulated as a dynamic fitness landscape and fourdifferent chaotic maps generating such a landscape are studied. By using acomputational approach, we analyze properties of the landscapes and quantifyt ...继续阅读 (4)

导读：文章分为两部分，Part1主要介绍金融数据分析的背景以及移动均线等方面的内容；Part2则介绍均线交叉策略的设计、回溯检验、基准测试以及实践中可能出现的若干问题，并结合Python代码实现了一个基于均线交叉的交易策略系统。当然文章最后还是提示大家仅仅是回溯检验预测的上涨并不意味着在未来会保持上涨的势头。所以：预测有风险，投资需谨慎。英文版：An Introduction to Stock Market Data Analysis with Python (Part 1)An Introduction to Stock Market Data Analysis with Python (Part 2)中文版：Python股市数据分析教程——学会它，或可以实现半“智能”炒股 (Part 1)Python股市数据分析教程——学会它，或可以实现半“智能”炒股 (Part 2) ...继续阅读 (2)

Information TheoryFinite Blocklength Analysis of Energy Harvesting ChannelsK Gautam Shenoy,Vinod SharmaComments: 18 pages, 1 FigureSubjects:Information Theory (cs.IT)We study DMC and AWGN channels when the transmitter harvests energy from theenvironment. These can model wireless sensor networks as well as Internet ofThings. We study such channels with infinite energy buffer in the finiteblocklength regime and provide the corresponding achievability and converseresults.The Matrix Exponential Distribution – A Tool for Wireless System Performance AnalysisPeter Larsson,Lars K. Rasmussen,Mik ...继续阅读 (5)

1. 我爱机器学习(2)[导读]Python股市数据分析教程arXiv Paper Daily: Wed, 21 Dec 20162. 机器之心(5)资源 | 基于MXNet 的神经机器翻译实现业界 | 既能打假也能造假，人工智能让图像和视频伪造更简单学界 | 谷歌新论文：使用生成对抗网络的无监督像素级域适应​专访 | 今日头条2016 Byte Cup大赛实战经验分享：要充分挖掘模型本身的信息重磅 | 美国白宫再发报告：三大策略应对人工智能驱动下的自动化经济3. 大数据文摘(3)工作职位推荐系统的算法与架构一人一笔 | 大数据行业从业者调研报告众筹全球最牛程序员出自哪些学校 中国“神秘”高中排名第二4. 36大数据(3)36早报 | GrowingIO打造国内最强数据分析，上海交大人工智能成功鉴别美女✪​ 洞察 | 2016年人们在搜什么？Google发布2016年热门关键词搜索榜单✪​ 实践 | 使用机器学习模型对大盘指数进行预测5. 深度学习世界(2)业界 | 既能打假也能造假，人工智能让图像和视频伪造更简单白宫发布《人工智能、自动化和经济》报告，敦促特朗普政府确保美国 AI 领先地位6. 爱可可 – 爱生活(29)《Nature 年度人物 Top 10：DeepMind CEO Hassabis 作为“Mind 缔造者”入选》via:@新智元 http://t ...继续阅读 (4)

Neural and Evolutionary ComputingA recurrent neural network without chaosThomas Laurent,James von BrechtSubjects:Neural and Evolutionary Computing (cs.NE); Computation and Language (cs.CL); Learning (cs.LG)We introduce an exceptionally simple gated recurrent neural network (RNN)that achieves performance comparable to well-known gated architectures, such asLSTMs and GRUs, on the word-level language modeling task. We prove that ourmodel has simple, predicable and non-chaotic dynamics. This stands in starkcontrast to more standard gated architectures, whose underlying dynamicalsystems exhibit cha ...继续阅读 (3)

1. 我爱机器学习(1)arXiv Paper Daily: Tue, 20 Dec 20162. 新智元(4)【邓侃】DeepMind 机器理解文本 NLP 技术复现与解析机器学习先驱 Michael I. Jordan 清华演讲：更好的分布式机器学习（PPT）Zuckerberg 亲写代码实现 AI 助理 Jarvis，跟 CEO 一起 100 小时实现小目标腾讯大数据推出第三代高性能计算平台 Angel，2017年全面开源3. 机器之心(5)深度 | 马克·扎克伯格展示个人助理Jarvis：又一个走出电影银幕的科幻作品学界 | 英特尔新论文提出事件驱动的随机反向传播：能实现神经形态深度学习业界 | 谷歌自动驾驶公司Waymo展示新车型，计划明年上路学界 | Andrej Karpathy：你为什么应该理解反向传播开源 | 哈佛大学NLP组开源神经机器翻译工具包OpenNMT：已达到生产可用水平4. 大数据文摘(3)视频更新｜斯坦福CS231n深度学习与计算机视觉课时17-迁移学习之物体定位与检测（下）这些神奇有趣的物理化学现象，你知道几个？IBM用机器学习探寻外星文明讯号 你也可以参与其中5. 36大数据(4)50张PPT：马云说未来将不再有电子商务，未来商业新常态是这样的当舆情大数据公司遭遇微博舆情如何破？清博大数据副总大骂被打记者遭网友批评投稿 | 决策树(专家药物模型) ...继续阅读 (5)

Neural and Evolutionary ComputingEvent-driven Random Back-Propagation: Enabling Neuromorphic Deep Learning MachinesEmre Neftci,Charles Augustine,Somnath Paul,Georgios DetorakisSubjects:Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI)An ongoing challenge in neuromorphic computing is to devise general andcomputationally efficient models of inference and learning which are compatiblewith the spatial and temporal constraints of the brain. The gradient descentbackpropagation rule is a powerful algorithm that is ubiquitous in deeplearning, but it relies on the immediate av ...继续阅读 (4)

1. 我爱机器学习(1)arXiv Paper Daily: Mon, 19 Dec 20162. 无痛的机器学习(1)番外篇(3)——最速下降法的特点3. 大数据文摘(3)想去机器学习初创公司做数据科学家？这里有最常问的40道面试题Nature | 美国研究所发布算法大赛 寻找可做癌症疫苗的蛋白突变D-News | 湖北大数据反腐颇见成效 雅虎又发生数据泄漏4. 36大数据(2)✪​ 教程 | 成为数据科学家的五个脚步✪​ 应用 | 金融行业大数据用户画像实践5. 深度学习世界(2)【预测＆盘点】深度学习热潮下，2017 年最受欢迎的编程语言是哪些【深度】“机器问题”卷土重来 人工智能要革谁的命？ ...继续阅读 (6)

Neural and Evolutionary ComputingGraphical RNN ModelsAshish Bora,Sugato Basu,Joydeep GhoshSubjects:Neural and Evolutionary Computing (cs.NE); Learning (cs.LG)Many time series are generated by a set of entities that interact with oneanother over time. This paper introduces a broad, flexible framework to learnfrom multiple inter-dependent time series generated by such entities. Ourframework explicitly models the entities and their interactions through time.It achieves this by building on the capabilities of Recurrent Neural Networks,while also offering several ways to incorporate domain knowledg ...继续阅读 (5)

Neural and Evolutionary ComputingStable Memory Allocation in the Hippocampus: Fundamental Limits and Neural RealizationWenlong Mou,Zhi Wang,Liwei WangSubjects:Neural and Evolutionary Computing (cs.NE); Data Structures and Algorithms (cs.DS); Learning (cs.LG)It is believed that hippocampus functions as a memory allocator in brain, themechanism of which remains unrevealed. In Valiant’s neuroidal model, thehippocampus was described as a randomly connected graph, the computation onwhich maps input to a set of activated neuroids with stable size. Valiantproposed three requirements for the hi ...继续阅读 (4)

Neural and Evolutionary ComputingUnderstanding the Impact of Precision Quantization on the Accuracy and Energy of Neural NetworksSoheil Hashemi,Nicholas Anthony,Hokchhay Tann,R. Iris Bahar,Sherief RedaComments: Accepted for conference proceedings in DATE17Subjects:Neural and Evolutionary Computing (cs.NE)Deep neural networks are gaining in popularity as they are used to generatestate-of-the-art results for a variety of computer vision and machine learningapplications. At the same time, these networks have grown in depth andcomplexity in order to solve harder problems. Given the limitations in ...继续阅读 (4)

1. 我爱机器学习(2)arXiv Paper Daily: Wed, 14 Dec 2016Machine Learning Daily: 2016-12-142. Hic Rhodus, hic salta(1)Owl of Minerva 的 Live －－ 科学管理压力与疲劳3. 新智元(5)深度学习训练时间从几小时缩减到数分钟，微软与CSCS超算新突破深度学习挑战冯·诺依曼结构【重磅】谷歌无人车成立独立实体公司 Waymo，从谷歌X实验室中剥离【首发】商汤科技获1.2亿美元新一轮融资，鼎晖、万达参投【快报】Facebook Oculus CEO 卸任 | AMD 进军 HPC 挑战英伟达4. 机器之心(5)业界 | 谷歌无人车项目成为独立公司Waymo，新车或2017年商用前沿 | 纽约大学发现新的言语工作记忆结构，可用于人工智能研究盘点 | 2016年不可错过的21个深度学习视频、教程和课程业界 | 微软开放小娜：推出两款开发工具提供软硬件支持资源 | 斯坦福大学NLP组开放神经机器翻译代码库（附论文）5. 大数据文摘(3)机器真的已经战胜人类医生了吗？医学AI标题党文章中的三大陷阱中科院陈润生院士: 精准医疗数据处理中的挑战一人一笔 | 大数据行业从业者调研报告众筹6. 36大数据(2)✪​ 洞察 | 2016年全球最轰动的七次数据泄漏事件(附数据安全企业名单)✪​ ...继续阅读 (4)

作者：Ross Fadely来源：https://blog.insightdatascience.com/原文链接：NIPS 2016 — Day 1 HighlightsNIPS 2016 — Day 2 Highlights: Platform wars, RL and RNNsNIPS 2016 — Day 3 Highlights: Robots that know, Cars that see, and more!NIPS 2016 — Final Highlights Days 4–6: Likelihood-free inference, Dessert analogies, and much more. ...继续阅读 (3)

Neural and Evolutionary ComputingNeurogenesis Deep LearningTimothy J. Draelos,Nadine E. Miner,Christopher C. Lamb,Craig M. Vineyard,Kristofor D. Carlson,Conrad D. James,James B. AimoneComments: Submitted to IJCNN 2017Subjects:Neural and Evolutionary Computing (cs.NE); Learning (cs.LG); Machine Learning (stat.ML)Neural machine learning methods, such as deep neural networks (DNN), haveachieved remarkable success in a number of complex data processing tasks. Thesemethods have arguably had their strongest impact on tasks such as image andaudio processing – data processing domains in which hu ...继续阅读 (21)

1. 我爱机器学习(3)NIPS 2016 — Daily HighlightsMachine Learning Daily: 2016-12-13arXiv Paper Daily: Tue, 13 Dec 20162. 无痛的机器学习(1)番外篇(2)——无聊的最速下降法推导3. 深度学习大讲堂专栏(2)深度学习框架Caffe源码解析【Technical Review】ECCV16 Grid Loss及其在人脸检测中的应用4. 新智元(1)The Information：谷歌8年无人车项目遭遇急刹车，或不再自主造车5. 机器之心(1)深度｜NIPS 2016最全盘点：主题详解、前沿论文及下载资源（附会场趣闻）6. 大数据文摘(3)视频更新｜斯坦福CS231n深度学习与计算机视觉课时16-迁移学习之物体定位与检测（上）Science | 分析20亿行北京出租车数据 寻找你被拒载的原因快讯 | 中国数据标准化及治理奖评选细则正式发布7. CSDN大数据(3)阿里超大规模Docker化之路MaxCompute SQL查询优化解析2017年深度学习十大趋势预测8. 36大数据(2)✪​ 投稿 | 如何不在互联网广告中花冤枉钱？✪​ 教程 | 如何思考个性化推荐的? ...继续阅读 (3)

来源：新智元(微信号AI_era)编辑：弗格森原文链接：图说2016深度学习十大指数级增长本文为新智元授权转载，严禁二次转载。【新智元导读】深度学习火热不是假象，以下的十个指数级增长趋势证明：深度学习在过去的一年内获得了极快的发展，已经成为当下人工智能的“顶梁柱”。本文从 ImageNet、NIPS和CVPR等顶级学术会议，以及谷歌、英伟达等企业应用，还有深度学习人才与投资、收购等角度分析，用13 张图解读深度学习到底有多火。1. 图像识别准确率的指数级增长似乎一切都是从 2015年的 ImageNet 挑战赛开始的，当年在图像识别准确率上，机器首次超过了人类，被认为是一个里程碑式的突破。图：ILSVRC top-5 错误率2010年算法的图像识别错误率至少在25%左右，但到2015年，计算机图像识别错误率已经低于人类（人类水平大概是4%左右）。2015年是0.03567，也就是3.5%。2016年，ImageNet 竞赛，图像识别错误率进一步下降，错误率今年的最好成绩为：平均错误率0.02991，也就是2.99%左右。从上图可以看到，在ImageNet 竞赛中，其实从2012年，开始，使用深度学习方法，错误率只有15.3 %左右。这是 Hinton 教授和他的两个研究生 Alex Krizhevsky, Illya Sutskever 的成果。据说，他们使用了 两个 Nvidia ...继续阅读 (4)

Neural and Evolutionary ComputingField-Programmable Crossbar Array (FPCA) for Reconfigurable ComputingMohammed A. Zidan,YeonJoo Jeong,Jong Hong Shin,Chao Du,Wei D. LuSubjects:Emerging Technologies (cs.ET); Hardware Architecture (cs.AR); Neural and Evolutionary Computing (cs.NE)For decades, advances in electronics were directly related to the scaling ofCMOS transistors according to Moore’s law. However, both the CMOS scaling andthe classical computer architecture are approaching fundamental and practicallimits, and new computing architectures based on emerging devices, such asnon-volatile ...继续阅读 (3)