Rasin Tsukuba Blog

A Brief Survey of Deep Reinforcement Learning

Algorithms and Challenges

A Brief Survey of Deep Reinforcement Learning Reinforcement Learning Algorithms There are two main approaches to solving RL problems: methods based on value functions and methods based on policy ...

Posted by Rasin on August 31, 2020

David Silver - Reinforcement Learning Note 2

Markov Decision Process

Lecture: Markov Decision Process Learning Goals Understand the Agent-Environment interface Understand what MDPs (Markov Decision Processes) are and how to interpret transition diagrams ...

Posted by Rasin on August 30, 2020

David Silver - Reinforcement Learning Note 1

Introduction What and Why

Lecture: Introduction to Reinforcement Learning About Reinforcement Learning Characteristics of Reinforcement Learning There is no supervisor, only a reward signal Feedback is delayed, ...

Posted by Rasin on August 29, 2020

Beginner's Guide to Deep Reinforcement Learning

Basic Knowledge

Origin: A Beginner’s Guide to Deep Reinforcement Learning Introduction Deep reinforcement learning combines artificial neural networks with a reinforcement learning architecture that enables soft...

Posted by Rasin on August 28, 2020

Learning to Navigate the Synthetiaclly Accessible Chemical Space Using Reinforcement Learning

Embedding the concept of synthetic accessibility directly into the drug design system

Abstract Current generative approaches exhibit a significant challenge as they do not ensure that the proposed molecular structures can be feasibly synthesized nor do they provide the synthesis ro...

Posted by Rasin on August 20, 2020

Introduction to GNN Note 3

Advanced Skills and Applications

Variants for Advanced Training Methods Sampling GCN requires the full-graph Laplacian, which is computational-consuming for large graphs. Futhermore, GCN is trained independently for a fixed grap...

Posted by Rasin on August 20, 2020

Introduction to GNN Note 2

Graph Neural Network and its Variations

Vanilla Graph Neural Networks We list the limitations of the vanilla GNN in representation capability and training efficiency. Introduction The concept of GNN aims to extend existing neural netw...

Posted by Rasin on August 19, 2020

Introduction to GNN Note 1

Basic Knowledge and Recap

Introduction Motivations Convolutional Neural Networks Firstly, GNNs are motivated by convolutional neural networks. We find the keys of CNNs: local connection shared weights use of mult...

Posted by Rasin on August 18, 2020

Graph Neural Network 1

Graph Neural Network Introduction

Origin: 图神经网络入门图神经网络入门图神经网络优势我们的推断主要来源于这个图的结构信息。对于深度模型如何有效使用这些图结构信息呢？经典的深度学习方法（比如DNN、RNN、CNN）能够有效的处理原始的输入，比如语音、图片、文本，但对于图结构信息，就不是很有效了。这样就引入了图神经网络。图神经网络的应用场景非常广泛，比如，社交网络（用户节点，关系为边）、电子购物（用户为和商品...

Posted by Rasin on August 13, 2020

Graph Embedding

Graph Embedding Introduction

Origin:图嵌入（Graph embedding）综述图嵌入综述图分析任务可以大致抽象为以下四类: 节点分类: 旨在基于其他标记的节点和网络拓扑来确定节点的标签(也称为顶点) 链接预测: 预测缺失链路或未来可能出现的链路的任务聚类: 聚类用于发现相似节点的子集，并将它们分组在一起可视化: 有助于深入了解网络结构。真实的图（网络）往往是高维、难以处理的...