Category Archives: deep learning

Special session on tensor methods for deep learning @ESANN2020

I am co-organizing with Danilo Mandic a special session on “Tensor Decompositions in Deep Learning” at ESANN 2020.

We welcome solid contributions and preliminary relevant results showing potential, limitations and challenges of new ideas related to the use of tensor decompositions in deep learning, neural networks, and machine learning at large.

Deadline for paper submission: 18 November 2019.

Prospective contributors/participants can contact me (or another co-organizer) for details.

Organized by Davide Bacciu (University of Pisa,Italy), Danilo Mandic (Imperial College, UK).

New Graduates

Congratulations to Luigi di Sotto and Stefan Daniel Motoc for having completed their M.Sc. and B.Sc. in Computer Science with a final project on deep learning topics. Luigi discussed a novel pooling mechanism for graph convolutional networks. Stefan proposed a thorough analysis of the latent space of MusAE, our deep adversarial autoencoder for music generation.

Paper accepted @IJCNN 2019

A paper on Bayesian tensor factorization for efficient processing of structured data has just been accepted for IJCNN 2019. Check it out:

Daniele Castellana, Davide Bacciu: Bayesian Tensor Factorisation for Bottom-up Hidden Tree Markov Models. Proceedings of the 2019 International Joint Conference on Neural Networks (IJCNN 2019I) , IEEE, 2019.

New ML graduates

Congratulations to two of my students who succesfully defended their B.Sc. and M.Sc. theses with two excellent ML projects.

Lorenzo Marsicano developed microESN, a library for the development of recurrent neural networks in embedded systems with minimal memory and computational resources.

Andrea Valenti designed a new deep learning model for MIDI music generation, beautifully named MusAE, for Music Adversarial autoEncoder (and of course to honour Greek mythology).

Congratulations!!

New IEEE TNNLS journal paper

Congratulations to my student Francesco Crecchi for his first journal paper showing how Dropout can be used to enforce robustness to missing inputs at test time in several recurrent neural networks. Check out its applications to sensor data processing.

Davide Bacciu, Francesco Crecchi : Augmenting Recurrent Neural Networks Resilience by Dropout. In: IEEE Transactions on Neural Networs and Learning Systems, 2019.

Paper accepted for ICRA 2019

A joint paper with the Hands and Haptics team at Centro Piaggio has just been accepted at the top-robotic confrence ICRA 2019 (and jointly to the IEEE Robotics and Automation Letters journal). A deep neural network learning, from humans, how to guide a robot arm in the manipulation of never-seen-before objects. Early access to the paper here: check out the upcoming videos of our system at work!

Della Santina Cosimo, Arapi Visar, Averta Giuseppe, Damiani Francesca, Fiore Gaia, Settimi Alessandro, Catalano Manuel Giuseppe, Bacciu Davide, Bicchi Antonio, Bianchi Matteo: Learning from humans how to grasp: a data-driven architecture for autonomous grasping with anthropomorphic soft hands. In: IEEE Robotics and Automation Letters, pp. 1-8, 2019, ISSN: 2377-3766, (Also accepted for presentation at ICRA 2019).

Accepted papers and sessions at ESANN’19

Good news in the Esann 2019 program!

Congratulations to Francesco Crecchi and Marco Podda for their accepted papers on adversarial attacks and graph generation

Francesco Crecchi, Davide Bacciu, Battista Biggio : Detecting Black-box Adversarial Examples through Nonlinear Dimensionality Reduction. Proceedings of the European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN'19), i6doc.com, Louvain-la-Neuve, Belgium, 2019.

Davide Bacciu, Alessio Micheli, Marco Podda: Graph generation by sequential edge prediction. Proceedings of the European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN'19), i6doc.com, Louvain-la-Neuve, Belgium, 2019.

Great success also for the special session on Societal Issues in Machine Learning: When Learning from Data is Not Enough I am co-organizing. We received an high number of great quality papers but only 4 made it to the oral plenary.

Good news on INNS BDDL 2019 program

The upcoming INNS Big Data and Deep Learning conference will be featuring two events from my group: a tutorial on Deep Learning for Graphs, jointly with Alessio Micheli, and an accepted oral paper survey deep learning models for tree transductions, jointly with my research associate Antonio Bruno.

Check out the INNS BDDL program here!

Bacciu Davide, Bruno Antonio: Deep Tree Transductions - A Short Survey. Proceedings of the 2019 INNS Big Data and Deep Learning (INNSBDDL 2019) , Recent Advances in Big Data and Deep Learning Springer International Publishing, 2019.

DeepDynamicHand paper now out

Check out my new joint work with Visar Arapi, Cosimo Della Santina, Matteo Bianchi and Antonio Bicchi. A deep learning based approach to recognize action primitives in video of object manipulation by humans. A first step towards automating robot manipulation skill acquisition from humans. 

Visar Arapi, Cosimo Della Santina, Davide Bacciu, Matteo Bianchi, Antonio Bicchi: DeepDynamicHand: A deep neural architecture for labeling hand manipulation strategies in video sources exploiting temporal information . In: Frontiers in Neurorobotics, vol. 12, pp. 86, 2018.