Category Archives: deep learning

Talk at Deep Learning Workshop in Trento

Here is a video of my recent talk on “Shaping Neural Networks with Dynamical Systems” at the DEEP LEARNING: Theory, Algorithms, and Applications 2023 Workshop in Trento.

In this 25min video I discuss the fundamentals of using dynamical systems’ concepts to imbue non-dissipation properties in neural networks and how we leverage dynamical systems as NNs building blocks in our EMERGE Project.

Video URL: https://drive.google.com/file/d/1RX8b-nCYqgR0sIOZLSkdrxDN35uLMG25/view

Best paper award @AAAI 2023

Great news from overseas as our paper on “Non-dissipative propagation by anti-symmetric deep graph networks” has just received the Best Student Paper Award 🏆 🍾 at the Deep Learning for Graphs workshop of AAAI23.

The paper is a great piece of work by Alessio Gravina, with a bit of support by Claudio Gallicchio and myself.

An extended version of it will s.oon be presented at ICLR 2023

New year updates

Quite a busy semester, hence long time no posts. Lets start the year with a bunch of good news.

Our book on Deep Learning in Biology and Medicine, edited with Paulo Lisboa, and Alfredo Vellido, is finally available for preorder. Check it out here!

Davide Bacciu, Paulo J. G. Lisboa, Alfredo Vellido: Deep Learning in Biology and Medicine. World Scientific Publisher, 2022, ISBN: 978-1-80061-093-4.

July 2022 is going to be a busy month in Padova with the organization of IEEE WCCI 2022, including

  • 1st Pervasive Artificial Intelligence Workshop co-chaired with Antonio Carta (Università di Pisa), Patrizio Dazzi (ISTI-CNR),  Magdalini Eirinaki (San Jose State University), Iraklis Varlamis (Harokopio University of Athens) – Deadline early April 2022
  • Deep learning for graphs special session, co-chaired with Shirui Pan (Monash University), Daniele Grattarola (IDSIA), Miao Zhang (Aalborg University), Nicolò Navarin (University of Padova), Feng Xia (Federation University Australia), Daniele Zambon (IDSIA) – Deadline 31st January 2022

IJCNN 2021 papers

Our group had 4 papers recently accepted to the upcoming International Joint Conference on Neural Networks! Much work on deep learning for graphs, including a novel edge-based model, an efficient graph generation approach and an explanation method for the chemical domain. Also a first proposal for an efficient federation of reservoir computing methods, part of our H2020 TEACHING efforts. Preprints soon on the Arxiv!

Daniele Atzeni, Davide Bacciu, Federico Errica, Alessio Micheli: Modeling Edge Features with Deep Bayesian Graph Networks. Proceedings of the International Joint Conference on Neural Networks (IJCNN 2021), IEEE IEEE, 2021.

Danilo Numeroso, Davide Bacciu: MEG: Generating Molecular Counterfactual Explanations for Deep Graph Networks. Proceedings of the International Joint Conference on Neural Networks (IJCNN 2021), IEEE 2021.

Davide Bacciu, Daniele Di Sarli, Pouria Faraji, Claudio Gallicchio, Alessio Micheli: Federated Reservoir Computing Neural Networks. Proceedings of the International Joint Conference on Neural Networks (IJCNN 2021), IEEE, 2021.

Davide Bacciu, Marco Podda: GraphGen-Redux: a Fast and Lightweight Recurrent Model for Labeled Graph Generation. Proceedings of the International Joint Conference on Neural Networks (IJCNN 2021), IEEE 2021.

Neurips 2020 WS papers

Excellent result by our group in the upcoming NeurIPS 2020 workshops with four accepted papers.
Congrats to Antonio Carta, Francesco Landolfi, Danilo Numeroso and Matteo Ronchetti!

Preprints coming up..

Matteo Ronchetti, Davide Bacciu: Generative Tomography Reconstruction. 34th Conference on Neural Information Processing Systems (NeurIPS 2020), Workshop on Deep Learning and Inverse Problems, 2020.

Davide Bacciu, Alessio Conte, Roberto Grossi, Francesco Landolfi, Andrea Marino: K-plex Cover Pooling for Graph Neural Networks. 34th Conference on Neural Information Processing Systems (NeurIPS 2020), Workshop on Learning Meets Combinatorial Algorithms, 2020.

Davide Bacciu, Danilo Numeroso: Explaining Deep Graph Networks with Molecular Counterfactuals. 34th Conference on Neural Information Processing Systems (NeurIPS 2020), Workshop on Machine Learning for Molecules - Accepted as Contributed Talk (Oral), 2020.

Antonio Carta, Alessandro Sperduti, Davide Bacciu : Short-Term Memory Optimization in Recurrent Neural Networks by Autoencoder-based Initialization . 34th Conference on Neural Information Processing Systems (NeurIPS 2020), Workshop on Beyond BackPropagation: Novel Ideas for Training Neural Architectures, 2020.

Paper Accepted at COLING 2020

Congratulations to Daniele Castellana for having his paper accepted at COLING 2020. Check it out if you are interested in higher-order neural networks for parse trees using tensor decompositions (soon on the Arxiv!).

Daniele Castellana, Davide Bacciu: Learning from Non-Binary Constituency Trees via Tensor Decomposition. PROCEEDINGS OF THE 2020 INTERNATIONAL CONFERENCE ON COMPUTATIONAL LINGUISTICS (COLING 2020), 2020.

New JMLR paper

Couldn’t think of a better venue for my 99th research paper than the Journal of Machine Learning Research. Check out our work on deep and probabilistic learning for graphs. Terrific job by Federico Errica!

Davide Bacciu, Federico Errica, Alessio Micheli: Probabilistic Learning on Graphs via Contextual Architectures. In: Journal of Machine Learning Research, vol. 21, no. 134, pp. 1−39, 2020.

Deep Learning for graph on Neural Networks

Very proud of the last effort from our group! Our tutorial paper on deep learning for graphs will be published as an invited paper on the Neural Networks journal!

Check out a preliminary version on the Arxiv!

Davide Bacciu, Federico Errica, Alessio Micheli, Marco Podda: A Gentle Introduction to Deep Learning for Graphs. In: Neural Networks, vol. 129, pp. 203-221, 2020.

COVID-19 Task Force

If you work in AI and you are willing to contribute to the worldwide fight against COVID-19 please consider joining the CLAIRE-COVID19 task force. I am coordinating the workgroup on omics, chemical and clinical data processing.

You can have a look at the first result of our pan-european collaboration which has been released here: it is a curated repository of protein-viral-drug-disease interactions for helping research in drug repurposing, bio-informatics, etc.