NicolaBernini / PapersAnalysis

Analysis, summaries, cheatsheets about relevant papers

21 stars 4 forks source link

Capsule Neural Networks for Graph Classification using Explicit Tensorial Graph Representations - ReadThrough #2

Open NicolaBernini opened 5 years ago

NicolaBernini commented 5 years ago

Original Paper

Capsule Neural Networks for Graph Classification using Explicit Tensorial Graph Representations

NicolaBernini commented 5 years ago

Possible Domain of Applications

Drug Discovery, Proteins, Chemicals, ...
Bioinformatics
Finance
Social Networks

NicolaBernini commented 5 years ago

Overview

Performing Graph Analysis, which means solving some Graph Specific Problems, can allow to solve more abstract problems related to Data Analysis

Graph Analysis Specific Tasks

Link Prediction
Node Classification

Related Higher Level Applications

Recommender Systems
Entity Disambiguation

NicolaBernini commented 5 years ago

Traditional Tools

Kernels

Kernel Methods

Pro

Work well in estimating Graph Similarity related to Graphs of different dimension

Con

Limited Scalability with respect to Graph Size
Some subclasses rely on mapping from variable to fixed size but
- different options (crop, resize, ...) and each can introduce artifacts
- in the context of graph there is the problem of preserving permutation