/*

Sed amet aliquam

Nullam et orci eu lorem consequat tincidunt vivamus et sagittis magna sed nunc rhoncus condimentum sem. In efficitur ligula tate urna. Maecenas massa vel lacinia pellentesque lorem ipsum dolor. Nullam et orci eu lorem consequat tincidunt. Vivamus et sagittis libero. Nullam et orci eu lorem consequat tincidunt vivamus et sagittis magna sed nunc rhoncus condimentum sem. In efficitur ligula tate urna.

Orci maecenas

Nullam et orci eu lorem consequat tincidunt vivamus et sagittis magna sed nunc rhoncus condimentum sem. In efficitur ligula tate urna. Maecenas massa sed magna lacinia magna pellentesque lorem ipsum dolor. Nullam et orci eu lorem consequat tincidunt. Vivamus et sagittis tempus.

Rhoncus magna

Nullam et orci eu lorem consequat tincidunt vivamus et sagittis magna sed nunc rhoncus condimentum sem. In efficitur ligula tate urna. Maecenas massa sed magna lacinia magna pellentesque lorem ipsum dolor. Nullam et orci eu lorem consequat tincidunt. Vivamus et sagittis tempus.

Sed nunc ligula

Nullam et orci eu lorem consequat tincidunt vivamus et sagittis magna sed nunc rhoncus condimentum sem. In efficitur ligula tate urna. Maecenas massa sed magna lacinia magna pellentesque lorem ipsum dolor. Nullam et orci eu lorem consequat tincidunt. Vivamus et sagittis tempus.

Massa libero

Nullam et orci eu lorem consequat tincidunt vivamus et sagittis libero. Mauris aliquet magna magna sed nunc rhoncus pharetra. Pellentesque condimentum sem. In efficitur ligula tate urna. Maecenas laoreet massa vel lacinia pellentesque lorem ipsum dolor. Nullam et orci eu lorem consequat tincidunt. Vivamus et sagittis libero. Mauris aliquet magna magna sed nunc rhoncus amet pharetra et feugiat tempus.

*/

Multiagent Bidirectionally-Coordinated Nets for Learning to Play StarCraft Combat Games

Multiagent Bidirectionally-Coordinated Nets for Learning to Play StarCraft Combat Games

Peng Peng (1), Quan Yuan (1), Ying Wen (2), Yaodong Yang (2), Zhenkun Tang (1), Haitao Long (1), Jun Wang (2)

(1) Alibaba Group, (2) University College London

(arXiv:1703.10069, Submitted on 29 Mar 2017)

Real-world artificial intelligence (AI) applications often require multiple agents to work in a collaborative effort

node2vec: Scalable Feature Learning for Networks

node2vec: Scalable Feature Learning for Networks

Aditya Grover, Jure Leskovec

Stanford University

adityag@cs.stanford.edu,jure@cs.stanford.edu

KDD ’16, August 13 - 17, 2016, San Francisco, CA, USA

Prediction tasks over nodes and edges in networks require careful effort in engineering features used by learning algorithms.

Awesome Papers: 2017-02-4

ACTOR-MIMIC: DEEP MULTITASK AND TRANSFER REINFORCEMENT LEARNING

Emilio Parisotto, Jimmy Ba, Ruslan Salakhutdinov

The ability to act in multiple environments and transfer previous knowledge to new

Awesome Papers: 2017-02-4

ACTOR-MIMIC: DEEP MULTITASK AND TRANSFER REINFORCEMENT LEARNING

Emilio Parisotto, Jimmy Ba, Ruslan Salakhutdinov

The ability to act in multiple environments and transfer previous knowledge to new

Awesome Papers: 2017-02-1

On The Construction of Extreme Learning Machine for Online and Offline One-Class Classification - An Expanded Toolbox

Chandan Gautam, Aruna Tiwari and Qian Leng

One-Class Classification (OCC) has been prime concern for researchers and effectively employed in various disciplines.

Awesome Papers: 2017-01-4

A K-fold Method for Baseline Estimation in Policy Gradient Algorithms

Nithyanand Kota, Abhishek Mishra, Sunil Srinivasa, Xi (Peter) Chen, Pieter Abbeel

The high variance issue in unbiased policy-gradient methods such as VPG and REINFORCE is typically mitigated by adding a baseline. However, the baseline fitting itself suffers from the underfitting or the overfitting problem. In this paper, we develop a K-fold method for baseline estimation in policy gradient algorithms.

Awesome Papers: 2017-01-3

Online Reinforcement Learning for Real-Time Exploration in Continuous State and Action Markov Decision Processes

Ludovic Hofer, Hugo Gimbert

This paper presents a new method to learn online policies in continuous state, continuous action, model-free Markov decision processes, with two properties that are crucial for practical applications.

Awesome Papers: 2017-01-2

Why and When Can Deep – but Not Shallow – Networks Avoid the Curse of Dimensionality: a Review

Tomaso Poggio, Hrushikesh Mhaskar, Lorenzo Rosasco, Brando Miranda, Qianli Liao

The paper reviews and extends an emerging body of theoretical results on deep learning including the conditions under which it can be exponentially better than shallow learning.

Awesome Papers: 2017-01-1

iCaRL: Incremental Classifier and Representation Learning

Sylvestre-Alvise Rebuffi, Alexander Kolesnikov, Christoph H. Lampert

A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data.

Awesome Papers: 2016-12-4

Deep Reinforcement Learning with Averaged Target DQN

Oron Anschel, Nir Baram, Nahum Shimkin

The commonly used Q-learning algorithm combined with function approximation induces systematic overestimations of state-action values. These systematic errors might cause instability, poor performance and sometimes divergence of learning.

Awesome Papers: 2016-12-3

Deep Convolutional Neural Network Design Patterns

Leslie N. Smith, Nicholay Topin

Recent research in the deep learning field has produced a plethora of new architectures. At the same time, a growing number of groups are applying deep learning to new applications.

Awesome Papers: 2016-12-2

Deep Unsupervised Learning using Nonequilibrium Thermodynamics

Jascha Sohl-Dickstein,Eric A Weiss,Niru Maheswaranathan, Surya Ganguli

A central problem in machine learning involves modeling complex data-sets using highly flexible families of probability distributions in which learning, sampling, inference, and evaluation are still analytically or computationally tractable.

Awesome Papers: 2016-12-1

Intriguing properties of neural networks

Christian Szegedy, Wojciech Zaremba,Ilya Sutskever,Joan Bruna, Dumitru Erhan,Ian Goodfellow, Rob Fergus

Deep neural networks are highly expressive models that have recently achieved state of the art performance on speech and visual recognition tasks. While their expressiveness is the reason they succeed, it also causes them to learn uninterpretable solutions that could have counter-intuitive properties. In this paper we report two such properties.

Awesome Papers: 2016-12-2

Bayesian Optimization for Machine Learning : A Practical Guidebook

Ian Dewancker, Michael McCourt, Scott Clark

The engineering of machine learning systems is still a nascent field; relying on a seemingly daunting collection of quickly evolving tools and best practices.

Awesome Papers: 2016-12-1

Intriguing properties of neural networks

Christian Szegedy, Wojciech Zaremba,Ilya Sutskever,Joan Bruna, Dumitru Erhan,Ian Goodfellow, Rob Fergus

Deep neural networks are highly expressive models that have recently achieved state of the art performance on speech and visual recognition tasks. While their expressiveness is the reason they succeed, it also causes them to learn uninterpretable solutions that could have counter-intuitive properties. In this paper we report two such properties.

Awesome Papers: 2016-11-1

Towards Deep Symbolic Reinforcement Learning

Deep reinforcement learning (DRL) brings the power of deep neural networks to bear on the generic task of trial-and-error learning, and its effectiveness has been convincingly demonstrated on tasks such as Atari video games and the game of Go. However, contemporary DRL systems inherit a number of shortcomings from the current generation of deep learning techniques. For example, they require very large datasets

Awesome Papers: 2016-09-1

Distributed Bayesian Learning with Stochastic Natural-gradient Expectation Propagation and the Posterior Server

Leonard Hasenclever, Stefan Webb, Thibaut Lienart, Yee Whye Teh, Sebastian Vollmer, Balaji Lakshminarayanan, Charles Blundell

Awesome Papers: 2016-09-2

Awesome Paper 2016-06-1

Input: Query and All related documents process:

  1. Tokenize all documents and query into a vectors. input: all the query and documents

Quantum Decision: A Survey

人类决策往往体现出非完全理性的特点,这是传统决策科学面临的巨大挑战之一,同时也是量子决策的主要研究对象。量子决策将决策的整个过程描述为认知过程。量子决策认为其背后的数学规律符合量子理论,因此量子决策只是将量子理论作为一种数学工具引入决策科学,其目的在于更好地描述和探索非完全理性这一特性,从而帮助人们更加深刻的认识决策过程,并不是认为大脑的工作过程符合量子理论。同时需要说明的是它并非是一种优化方法。 也就是说,它的初衷并非寻找一个最优策略,而是预测在某种环境下人们最可能做出何种选择。目前量子决策还处于非常稚嫩的阶段,往往用在经济、博弈论和认知科学中,而且应用相对比较简单。

Modeling and Simulation Research on Propagation of Public Opinion

Considering nowadays propagation of Public Opinion possessing different regular pattern and effect, it is the respond of reality demand to research on network ropagation mechanism of public opinion. Simulation is conducted based on network structured by improved scale-free network. Instead of study propagation by single method, the paper analyzes the view exchange process in detail by introducing different degree of individual’s stubbornness and belief degree. Deffaunt Model is used and improved. By studying two kinds of initialization, the paper finds out people will have distinct opinion after propagation in the context of non-intervention, learning the nature of network’s influence in propagation. Furthermore, when intervention is taken into

OCC or PU Learning2.0

OCC or PU Learning Definition: One Class Classification or Positive and Unlabeled Learning corresponds to a special case in machine learning when only a small proportion of positive samples and unlabeled ones are employed for learning task. One frequent task is generally to retrieve positive ones from the unlabeled in the training dataset, namely transductive learning. Detecting positive or identifying negative ones from the unlabeled in both training dataset and unknown is another popular task, also referred to as inductive learning.

Drug Target Protein-Protein Interaction Networks: A Systematic Perspective

The identification and validation of drug targets is crucial in biomedical research. Most recently, lots of studies have been conducted on analyzing drug target features aiming at getting a better understanding on principles of their mechanisms. But most of them are based on either strong biological hypotheses or the chemical and physical properties of those targets separately. In this paper, we examined three main ways to understand the functional biomolecules based on the topological features of drug targets. By comparison between two types of proteins (targets and common proteins) of the protein-protein interactions (PPI) network, the results show there are no significant differences on intermediary and source functions.

Efficient Data Mining Algorithms for Screening Potential Proteins of Drug Targets

The past decades of years has witnessed the booming of pharmacology as well as the dilemma of drug development. Playing in a crucial role in drug design, retrieval of potential proteins of drug targets from open access database is an overwhelming task of challenge but significance, receiving much attention from areas of academia and medical engineering. Human proteins with several measured physical-chemical properties can be well mined for screening reliable candidates, which would accelerate the process of drug development and bring the reduction in expense. In this paper, retrieval of potential drug target proteins (DTPs) from a fine collected dataset was researched from the perspective of data mining. Puzzled with the uncertainty of non-drug target proteins (NDTPs),

A Novel Ensemble Method for Imbalanced Data Learning-Bagging of Extrapolation-SMOTE SVM

One class classification (OCC) problem corresponds to a special case in machine learning area when only a proportion of positive samples and unlabeled ones are employed for learning tasks. Due to the similar characteristics of the learning dataset, we also recognize the positive and learning problems (PU) as equivalent to the OCC problems in this paper. At the same time, some other concepts such as anomaly detection, novelty detection, concept learning show great connection with the OCC when it comes to some specific task. Obviously, the OCC behaves significantly different from the traditional supervised learning paradigm in which the former neg-lects the prior distribution of classes and gets rid of the dependence on the negative

Dynamic Fault detection and Optimization of Assignment for Decision-makers in C2-Organization

The Command and Control (C2) organization plays an indispensable role on planning of military operation. Nowadays the plan oriented more uncertaintybattlefield environment is the hot area. Many studies highlight how to rapidlymake or change the plans with detecting advance information. One of the solutions is based the time-domain metric model to evaluate C2 organizational decision-making capability. We develop an improved simulated annealing algorithm to facilitate this model implementation. In order todiscretize the time-domain, we propose a horizon partition that is based on the task dynamic state.Finally, the optimization

Peeling the Drug Target Proteins Network by a Core Decomposition Method

Currently as the emergence of high throughput technologies for omics data, such as yeast 2 hybrid protein interactions, researches of protein-protein interaction (PPI) had been explosively conducted on. One of the consensuses is that topographical analysis of the intercellular protein interactions lead to new avenues for drug target prediction. In this paper, a panorama view of drug targets is constructed. We found that the connectivity is often not a sufficient criterion to analyze the function of the drug target. On the other hand, the core of the PPI network poses a systematic way to consider the local and global significance of a protein, which indicates the inherent layer structure of the targets interactions.

Drug Targets Properties Analysis and Prediction with Machine Learning

Motivation: Identifying new drug protein plays an important role in pharmaceutical and biomedical research. Before new drug target get manual tested and put into experimental study, a globe filter method is more efficient, practical and economical. New methods are developed to predict new drug targets, but most of them are hypothesis specific or data specific, thus, a more general computational method is required. Result: In this study, we first category the target protein as carrier, transporter, and enzyme, gained 153 overlap proteins among these three categories which should be further analyzed. Using their sequence information we gained their

Drug target proteins prediction with network properties

Motivation: The identification of drug target proteins can certainly benefit new drug development. With the advances in high throughput sequencing technologies, proteins are studied systematically as a network namely protein-protein interaction (PPI) network. PPI plays a fundamental role in many biological processes. Therefore, it provides us a better way to understand drug and protein interaction. In this paper, we analyzed the drug target protein’s interaction network and extracted the physical and chemical properties from protein sequence information. Based both of the network and individual information, the potential drug target proteins are predicted to help identify new drug targets.