In this intensive course, advanced topics in the wide range of informatics such as mathematical information sciences, intelligence sciences, life-sciences and socio-economic sciences are introduced by visiting lecturers.
The aim of this course is to broaden students' perspectives by lectures of advanced topics by active scientists in the front line.
advanced topics in wide range of informatics such as mathematical information sciences, intelligence sciences, life-sciences and socio-economic sciences are introduced by visiting lecturers.
The aim of this course is to broaden students' perspective by lectures of advanced topics by active scientists in the front line.
In the first 7 lectures, a wide variety of things and their relationships in the digital and physical world can be represented as graph. In the first 7 lectures, we study deep learning based methods called "graph neural networks or GNNs" that enable representation learning on graph-structured data. We also learn how GNNs can be used for real-world applications such as recommender systems, anomaly detection in financial institutions, material discovery, and so forth.
In the second half of lecture series, we present Markov chain Monte Carlo (MCMC) methods and closely related stochastic algorithms. We begin our discussion with the review of Markov chains and random algorithms in a general setting, preparing the stage for the study of various implementations of stochastic algorithms. We explore other interesting topics such as hidden Markov models, Langevin algorithms, and applications in Bayesian statistics.
Students can obtain knowledge about advanced topics in mathematical information sciences, intelligence sciences, life sciences and socio-economic sciences.
mathematical information sciences, intelligence sciences, life sciences, socio-economic sciences
✔ Specialist skills | Intercultural skills | Communication skills | ✔ Critical thinking skills | Practical and/or problem-solving skills |
Lectures give intensive lectures about selected advanced topics.
Course schedule | Required learning | |
---|---|---|
Class 1 | Graph structures and graph analytics for real-world applications | Graph theory |
Class 2 | Methods to understand graph structures | Graph theory |
Class 3 | Introduction to Graph Neural Networks | Deep Learning |
Class 4 | Graph Neural Networks for Heterogenous Graphs | Deep Learning |
Class 5 | Graph Neural Networks for Time-Evolving Graphs | Deep Learning |
Class 6 | Graph Neural Networks for Recommender Systems | Deep Learning |
Class 7 | State-of-the-art research themes on Graph Neural Networks | Deep Learning |
Class 8 | Motivation for Monte Carlo simulation / Rejection algorithm | Review for probabilistic approach and Monte Carlo simulation |
Class 9 | Markov chain Monte Carlo method / Metropolis algorithm | Markov chains and sampling algorithm |
Class 10 | Discrete structure and Gibbs sampler / Gibbs algorithm | Gibbs model and sampling algorithm |
Class 11 | How long should you run it? / Perfect sampling algorithms | Coupling and perfect sampling methods |
Class 12 | Hidden Markov model and dynamic decision making / Viterbi algorithm | Hidden Markov model |
Class 13 | Brownian motion and Monte Carlo simulation / Langevin algorithm | Brownian motion and sampling algorithm |
Class 14 | MCMC in practice / Bayesian computation | Bayesian methods and simulation |
None
Specified by lecturers
Will be based on exercise and report.
None
MUROFUSHI, Toshiaki (murofusi[at]c.titech.ac.jp)
The details will be announced later.