《nature》 机器学习的3个热门分支 “深度学习”、“增强式学习”、“概率学习” 的最新综述

《nature》 机器学习各个分支（branch of machine learning） “深度学习”、“增强式学习”、“概率学习” 的最新的综述。等有空了 把他们翻译成中文

1.Deep learning，Yann LeCun, Yoshua Bengio & Geoffrey Hinton ，2 8 M AY 2 0 1 5 | VO L 5 2 1 | N AT U R E | 深度学习

   Abstract：Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition,visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.

 2.Reinforcement learning improves behaviour from evaluative feedback. Michael L. Littman.2 8 M AY 2 0 1 5 | VO L 5 2 1 | N AT U R E | ,增强式学习

 Abstract:Reinforcement learning is a branch of machine learning concerned with using experience gained through interacting with the world and evaluative feedback to improve a system’s ability to make behavioural decisions. It has been called the artificial intelligence problem in a microcosm because learning algorithms must act autonomously to perform well and achieve their goals. Partly driven by the increasing availability of rich data, recent years have seen exciting advances in the theory and practice of reinforcement learning, including developments in fundamental technical areas such as generalization,planning, exploration and empirical methodology, leading to increasing applicability to real-life problems.

3.Probabilistic machine learning and artificial intelligence.Zoubin Ghahramani.2 8 M AY 2 0 1 5 | VO L 5 2 1 | N AT U R E ,“概率学习”

Abstract: How can a machine learn from experience? Probabilistic modelling provides a framework for understanding what learning is, and has therefore emerged as one of the principal theoretical and practical approaches for designing machines that learn from data acquired through experience. The probabilistic framework, which describes how to represent and manipulate uncertainty about models and predictions, has a central role in scientific data analysis, machine learning,robotics, cognitive science and artificial intelligence. This Review provides an introduction to this framework, and discusses some of the state-of-the-art advances in the field, namely, probabilistic programming, Bayesian optimization,data compression and automatic model discovery.

此外nature 上面也有一篇 “深度学习”结合“增强式学习”的paper-------human level control through deep reinforcement learning（DRL）,DeepMind's Nature Paper and Earlier Related Work对该paper的前前后后发展介绍的网页：http://people.idsia.ch/~juergen/naturedeepmind.html