Posts by Collection

portfolio

Lifted Relational Neural Networks

A custom framework for deep relational learning, written completely from scratch in Java (with a Python frontend)
see my NeuraLogic repository
date: 2015-2020
role: developer

End-to-end learning of optimal portfolios

Focus on principled combining of machine learning and subsequent (portfolio) optimization problems.
date: 2020-present
role: lead researcher
funding: Czech Science Foundation

Deep Relational Learning

Developing novel algorithms for transferring neural network methods into relational learning setting.
date: 2017-2019
role: lead researcher
funding: Czech Science Foundation

Machine learning for network security

Developing machine learning methods and sound algorithms to detect malicious events from NetFlow traffic.
date: 2013-2015
role: researcher
funding: Cisco R&D

Machine learning methods for bio-informatics

Developing various novel machine learning methods for data analysis in bio-informatics (and beyond).
date: 2017-present
role: coordinator (PI)
funding: Czech Technical University

publications

Lifted Relational Neural Networks: Efficient Learning of Latent Relational Structures

Published in Journal of Artificial Intelligence Research, 2015

This paper introduces our original concept of Lifted Relational Neural Networks (LRNN, from 2015)

Download here

Events from network flows

Published in US patent office, 2016

US patent - A constructive procedure for detection of malicious network events from NetFlow traffic.

Download here

Stacked Structure Learning for Lifted Relational Neural Networks

Published in International Conference on Inductive Logic Programming, 2017

A structure learning extension of the LRNN concept - we won the best paper of the conference with it.

Download here

Beyond Graph Neural Networks with Lifted Relational Neural Networks

Published in Machine Learning Journal, 2020

We show how the Lifted Relational Neural Networks concept covers the Graph Neural Networks concept (and more)

Download here

Beating the market with a bad predictive model

Published in International Journal of Forecasting, 2020

I wrote this paper to closely explain our previous concept of exploiting sports betting market with ML

Download here

Lossless compression of structured convolutional models via lifting

Published in International Conference on Learning Representations (ICLR), 2021

We introduce a lossless compression technique to significantly speed up models such as Graph Neural Networks

Download here

Deep Learning with Relational Logic Representations

Published in Czech Technical University press (theses), 2021

This is actually my dissertation thesis - it contains my work on deep relational learning, which got also published as a book

Download here

Gustav Šír

Posts by Collection

portfolio

Lifted Relational Neural Networks

End-to-end learning of optimal portfolios

Deep Relational Learning

Machine learning for network security

Machine learning methods for bio-informatics

publications

Lifted Relational Neural Networks: Efficient Learning of Latent Relational Structures

Events from network flows

Stacked Structure Learning for Lifted Relational Neural Networks

Beyond Graph Neural Networks with Lifted Relational Neural Networks

Beating the market with a bad predictive model

Lossless compression of structured convolutional models via lifting

Deep Learning with Relational Logic Representations

talks

Talk 1 on Relevant Topic in Your Field

Tutorial 1 on Relevant Topic in Your Field

Talk 2 on Relevant Topic in Your Field

Conference Proceeding talk 3 on Relevant Topic in Your Field

teaching

Functional and Logic Programming

Introduction to Natural Language Processing

Machine Learning and Data Analysis

Database Systems

Symbolic Machine Learning