本周,机器学习和计算神经科学领域的顶级大会第 30 届国际神经信息处理系统大会(NIPS2016)在巴塞罗那举办,内容包括演讲、展示和宣讲和海报展示,在这里可以一睹最新的机器学习研究。谷歌带着 280 名员工强势亮相,除了技术演讲和海报展示外,他们还将举办研讨会和多个 tutorials。
谷歌的研究一直走在机器学习的前沿,积极探索机器学习的各个方面,包括经典算法以及像深度学习这样的前沿技术,既关注理论也重视应用。他们在语言理解、语音、翻译、视觉处理、排名和预测上的很多成果都依赖于机器智能。在所有的任务中,他们收集了大量直接或间接的利益关系的证据,并开发学习理解和泛化的方法。
Invited Talk
标题:Dynamic Legged Robots
作者:Marc Raibert
新一代高性能的机器人正在离开实验室进入现实世界,出现在办公室、家庭以及一些普通机器无法到达的地方。这些新兴机器人使用探测器来观察周边,并依靠其在环境中导航,理解环境,与环境互动。它们敏捷、灵巧和自主和智能都在按照将人类从各种任务中解放出来的愿景在不断发展进化。
28篇论文
1.论文:Boosting with Abstention
作者:Corinna Cortes, Giulia DeSalvo, Mehryar Mohri
论文地址:http://papers.nips.cc/paper/6335-boosting-with-abstention
2.论文: Community Detection on Evolving Graphs
作者:Stefano Leonardi, Aris Anagnostopoulos, Jakub Łącki, Silvio Lattanzi, Mohammad Mahdian
论文地址:http://papers.nips.cc/paper/6173-community-detection-on-evolving-graphs.pdf
3.论文:Linear Relaxations for Finding Diverse Elements in Metric Spaces
作者:Aditya Bhaskara, Mehrdad Ghadiri, Vahab Mirrokni, Ola Svensson
论文地址:http://papers.nips.cc/paper/6500-linear-relaxations-for-finding-diverse-elements-in-metric-spaces.pdf
4.论文:Nearly Isometric Embedding by Relaxation
作者:James McQueen, Marina Meila, Dominique Joncas
论文地址:http://papers.nips.cc/paper/6535-nearly-isometric-embedding-by-relaxation.pdf
5.论文:Optimistic Bandit Convex Optimization
作者:Mehryar Mohri, Scott Yang
论文地址:http://papers.nips.cc/paper/6429-optimistic-bandit-convex-optimization.pdf
6.论文:Reward Augmented Maximum Likelihood for Neural Structured Prediction
作者:Mohammad Norouzi, Samy Bengio, Zhifeng Chen, Navdeep Jaitly, Mike Schuster, Yonghui Wu, Dale Schuurmans
论文地址:http://papers.nips.cc/paper/6547-reward-augmented-maximum-likelihood-for-neural-structured-prediction.pdf
7.论文:Stochastic Gradient MCMC with Stale Gradients
作者:Changyou Chen, Nan Ding, Chunyuan Li, Yizhe Zhang, Lawrence Carin
论文地址:http://papers.nips.cc/paper/6359-stochastic-gradient-mcmc-with-stale-gradients.pdf
8.论文:Unsupervised Learning for Physical Interaction through Video Prediction
作者:Chelsea Finn*, Ian Goodfellow, Sergey Levine
论文地址:http://papers.nips.cc/paper/6161-unsupervised-learning-for-physical-interaction-through-video-prediction.pdf
9.论文:Using Fast Weights to Attend to the Recent Past
作者:Jimmy Ba, Geoffrey Hinton, Volodymyr Mnih, Joel Leibo, Catalin Ionescu
论文地址:http://papers.nips.cc/paper/6057-using-fast-weights-to-attend-to-the-recent-past.pdf
10.论文:A Credit Assignment Compiler for Joint Prediction
作者:Kai-Wei Chang, He He, Stephane Ross, Hal III
论文地址:http://papers.nips.cc/paper/6256-a-credit-assignment-compiler-for-joint-prediction.pdf
11.论文:A Neural Transducer
作者:Navdeep Jaitly, Quoc Le, Oriol Vinyals, Ilya Sutskever, David Sussillo, Samy Bengio
论文地址:http://papers.nips.cc/paper/6594-an-online-sequence-to-sequence-model-using-partial-conditioning.pdf
12.论文:Attend, Infer, Repeat: Fast Scene Understanding with Generative Models
作者:S. M. Ali Eslami, Nicolas Heess, Theophane Weber, Yuval Tassa, David Szepesvari, Koray Kavukcuoglu, Geoffrey Hinton
论文地址:http://papers.nips.cc/paper/6230-attend-infer-repeat-fast-scene-understanding-with-generative-models.pdf
13.论文:Bi-Objective Online Matching and Submodular Allocations
作者:Hossein Esfandiari, Nitish Korula, Vahab Mirrokni
论文地址:http://papers.nips.cc/paper/6085-bi-objective-online-matching-and-submodular-allocations.pdf
14.论文:Combinatorial Energy Learning for Image Segmentation
作者:Jeremy Maitin-Shepard, Viren Jain, Michal Januszewski, Peter Li, Pieter Abbeel
论文地址:http://papers.nips.cc/paper/6595-combinatorial-energy-learning-for-image-segmentation.pdf
15.论文:Deep Learning Games
作者:Dale Schuurmans, Martin Zinkevich
论文地址:http://papers.nips.cc/paper/6315-deep-learning-games.pdf
16.论文:DeepMath - Deep Sequence Models for Premise Selection
作者:Geoffrey Irving, Christian Szegedy, Niklas Een, Alexander Alemi, François Chollet, Josef Urban
论文地址:http://papers.nips.cc/paper/6280-deepmath-deep-sequence-models-for-premise-selection.pdf
17.论文:Density Estimation via Discrepancy Based Adaptive Sequential Partition.
作者:Dangna Li, Kun Yang, Wing Wong
论文地址:http://papers.nips.cc/paper/6217-density-estimation-via-discrepancy-based-adaptive-sequential-partition.pdf
18.论文:Domain Separation Networks
作者:Konstantinos Bousmalis, George Trigeorgis, Nathan Silberman, Dilip Krishnan, Dumitru Erhan
论文地址:http://papers.nips.cc/paper/6254-domain-separation-networks.pdf
19.论文:Fast Distributed Submodular Cover: Public-Private Data Summarization
作者:Baharan Mirzasoleiman, Morteza Zadimoghaddam, Amin Karbasi
http://papers.nips.cc/paper/6540-fast-distributed-submodular-cover-public-private-data-summarization.pdf
20.论文:Satisfying Real-world Goals with Dataset Constraints
作者:Gabriel Goh, Andrew Cotter, Maya Gupta, Michael P Friedlander
论文地址:http://papers.nips.cc/paper/6316-satisfying-real-world-goals-with-dataset-constraints.pdf
21.论文:Can Active Memory Replace Attention?
作者:Łukasz Kaiser, Samy Bengio
论文地址:http://papers.nips.cc/paper/6295-can-active-memory-replace-attention.pdf
22.论文:Fast and Flexible Monotonic Functions with Ensembles of Lattices
作者:Kevin Canini, Andy Cotter, Maya Gupta, Mahdi Fard, Jan Pfeifer
论文地址:http://papers.nips.cc/paper/6377-fast-and-flexible-monotonic-functions-with-ensembles-of-lattices.pdf
23.论文:Launch and Iterate: Reducing Prediction Churn
作者:Quentin Cormier, Mahdi Fard, Kevin Canini, Maya Gupta
论文地址:http://papers.nips.cc/paper/6053-launch-and-iterate-reducing-prediction-churn.pdf
24.论文:On Mixtures of Markov Chains
作者:Rishi Gupta, Ravi Kumar, Sergei Vassilvitskii
论文地址:http://papers.nips.cc/paper/6078-on-mixtures-of-markov-chains.pdf
25.论文:Orthogonal Random Features
作者:Felix Xinnan Yu, Ananda Theertha Suresh, Krzysztof Choromanski, Dan Holtmann-Rice, Sanjiv Kumar
论文地址:http://papers.nips.cc/paper/6246-orthogonal-random-features.pdf
26.论文:Perspective Transformer Nets: Learning Single-View 3D Object Reconstruction without 3D Supervision
作者:Xinchen Yan, Jimei Yang, Ersin Yumer, Yijie Guo, Honglak Lee
27.论文:Structured Prediction Theory Based on Factor Graph Complexity
作者:Corinna Cortes, Vitaly Kuznetsov, Mehryar Mohri, Scott Yang
论文地址:http://papers.nips.cc/paper/6485-structured-prediction-theory-based-on-factor-graph-complexity.pdf
28.论文:Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivity
作者:Amit Daniely, Roy Frostig, Yoram Singer
论文地址:http://papers.nips.cc/paper/6427-toward-deeper-understanding-of-neural-networks-the-power-of-initialization-and-a-dual-view-on-expressivity.pdf
Demonstrations
标题:Interactive musical improvisation with Magenta
作者:Adam Roberts, Sageev Oore, Curtis Hawthorne, Douglas Eck
我们结合了基于LSTM的循环神经网络和Deep Q-learning建立了实时生成音乐序列。LSTM的任务是学习音乐评分(编码为MIDI,而不是音频文件)的一般结构。Deep Q-learning用来改进基于奖励的序列,如期望的类型,组成正确性和预测人类协作者演奏的内容。基于RNN模型的生成与强化学习的结合是一种生成音乐的全新方式。这种方式比单独使用LSTM更为稳定,生成的音乐更加好听。该方法有两个任务:生成对短旋律输入的响应,以及实时生成对旋律输入的伴奏,持续对未来输出进行预测。本方法在TensorFlow中加入了一个全新的MIDI接口产生即兴的音乐体验,让使用者可以与神经网络实时交互。
标题:Content-based Related Video Recommendation
作者:Joonseok Lee
这是一个相关视频推荐的展示,种子来源于YouTube上随机的视频,纯粹基于视频内容信号。传统的推荐系统使用协同过滤(CF) 方法,在有多少用户在看了种子视频之后观看特定的候选视频的基础之上来推荐相关视频。这种方式没有考虑视频内容但是考虑了用户行为。在这个展示中,我们关注的是冷启动问题,其中种子或者候选视频都是新上传的(或者未被发现的)。对此我们按照一个基于视频内容的相似性学习问题进行建模,并学习了深度视频嵌入经过训练去预测真实情况的视频关系(由一个CF基于协同手表的系统鉴定) ,但仅使用视觉内容。它基于任一新视频内容,将其嵌入进一个1024维的表征中,同时成对视频的相似性在嵌入的空间中仅当做一个点积来计算。我们发现,被学习的视频嵌入超越了简单的视觉相似性,并能捕捉复杂的语义关系。
更多的 workshops 和 tutorials 可点击网址:https://research.googleblog.com/2016/12/nips-2016-research-at-google.html