Federico Errica

Federico Errica

PhD in Computer Science

NEC Laboratories Europe GmbH

About Me

Federico Errica received his PhD in Computer Science from the University of Pisa, supervised by Alessio Micheli and Davide Bacciu. He is now a Research Scientist at NEC Laboratories Europe GmbH. His research interests include deep probabilistic models for graphs, neural networks and hybrid architectures.

Download: CV.

Download: PhD thesis.

Download: PhD defense's slides.

Interests
  • Machine Learning for Graphs
  • Probabilistic Modeling
  • Hybrid Architectures
Education

  • PhD in Computer Science, 2018-2022

    University of Pisa

  • MSc in Computer Science, 2015-2018

    University of Pisa

  • BSc in Computer Science, 2012-2015

    University of Pisa

Experience

 
 
 
 
 
Research Scientist
NEC Laboratories Europe GmbH
Jan 2022 – Present Heidelberg, DE
 
 
 
 
 
Visiting Researcher
University College London
Jan 2021 – May 2021 London, UK
Worked on an empirical study of neural link predictors for node classification.
 
 
 
 
 
Research Intern
Facebook AI Research
Jun 2019 – Sep 2019 London, UK
Worked on a low-resource classificator for text.
 
 
 
 
 
Machine Learning Researcher
Laife Reply
Mar 2018 – Sep 2018 Milan, Italy
Worked on breast cancer detection.

Recent Posts

Paper accepted at ICML 2022
After a year of hard work, rejections, good feedback, paper rewriting, and additional experiments, our Infinite Contextual Graph Markov Model has been accepted at ICML 2022! iCGMM combines graph learning and Bayesian nonparametric to build a deep model that can decide the complexity of each of its layers and automatize the choice of its hyper-parameters during training.
Federico Errica
May 15, 2022 1 min read
Paper accepted at ICML 2022
PyDGN 1.0.0
PyDGN 1.0.0 is out! We have added a documentation, refactored the code, implemented a bunch of user-friendly features and provided clean, well commented configuration files!
Federico Errica
Mar 2, 2022 1 min read
PyDGN 1.0.0
Started a new job
While I prepare my Ph.D. thesis' defense, I have just started working for NEC Laboratories Europe as a Research Scientist! The people and the environment are truly stimulating, stay tuned for more updates on deep learning for graphs =).
Federico Errica
Jan 13, 2022 1 min read
PyDGN v0.5.0 is out!
PyDGN v0.5.0 is out
Federico Errica
Aug 13, 2021 1 min read
PyDGN v0.5.0 is out!
GMDN got accepted at ICML'21!
Deep Bayesian Graph Networks & Hybrid Methods
Federico Errica
May 9, 2021 1 min read
GMDN got accepted at ICML'21!
See all posts

Projects

Notes on the Reparametrization Trick
One of the most popular articles of the last years in the Machine Learning community is the Auto-Encoding Variational Bayes, which also includes the so-called reparametrization trick. Intrigued by what was sketched in the article, I decided to work out the details of this reparametrization, covering 2 of the 3 cases described (but I guess the third one can be derived from the first 2).
Notes on the Reparametrization Trick
Derivations of SNE and t-SNE
Recently I was looking at Stochastic Neigbhor Embedding (SNE) and its t-distributed version (t-SNE), but I could not find the exact steps to derive the gradient of the loss function (there are small errors in the t-SNE article and no info in the SNE one), so I decided to carry on the derivation and share it.
Derivations of SNE and t-SNE
PyDGN
A Python library to easily experiment with Deep Graph Networks.
PyDGN

Recent & Upcoming Talks

NEC Labs Europe
I talked about my PhD research
Jul 21, 2021 Virtual
Continual AI
I talked about our latest work on catastrophic forgetting for DGNs.
Apr 9, 2021 Virtual
IBM Research Zurich
I talked about our latest research on deep learning for graphs.
Apr 6, 2021 Virtual

Featured Publications

Daniele Castellana, Federico Errica, Davide Bacciu, Alessio Micheli
May 2022 ICML 2022
The Infinite Contextual Graph Markov Model

Combining Deep Graph Networks with Bayesian nonparametric techniques to unsupervised learning of node/graph representations while automatizing the choice of the hyper-parameters during training.
PDF Cite Code
Luca Oneto, Nicolo Navarin, Battista Biggio, Federico Errica, Alessio Micheli, Franco Scarselli, Monica Bianchini, Luca Demetrio, Pietro Bongini, Armando Tacchella, Alessandro Sperduti
April 2022 In Neurocomputing
Towards Learning Trustworthily, Automatically, and with Guarantees on Graphs: an Overview

An overview of the current works focused towards learning trustworthily, automatically, and with guarantees on graphs.
PDF Cite DOI
Federico Errica, Davide Bacciu, Alessio Micheli
May 2021 ICML 2021
Graph Mixture Density Networks

Combining Deep Graph Networks with Mixture Density Networks to model multimodal output distributions conditioned on arbitrary input graphs.
PDF Cite Code
Federico Errica, Marco Giulini, Davide Bacciu, Roberto Menichetti, Alessio Micheli, Raffaello Potestio
February 2021 In Frontiers Molecular Biosciences
A deep graph network-enhanced sampling approach to efficiently explore the space of reduced representations of proteins

Learning complex measure of information loss in reduce representations of proteins via deep graph networks.
PDF Cite Code DOI
Antonio Carta, Andrea Cossu, Federico Errica, Davide Bacciu
January 2021 The Web Conference 2021, Workshop on Graph Learning Benchmarks
Catastrophic Forgetting in Deep Graph Networks: an Introductory Benchmark for Graph Classification

An empirical work at the intersection of Continual Learning and Graph Representation Learning. Selected as one of the two oral presentations.
PDF Cite Code
Davide Bacciu, Federico Errica, Alessio Micheli
July 2020 In JMLR
Probabilistic Learning on Graphs via Contextual Architectures

Further studies on the Contextual Graph Markov Model.
PDF Cite Code
Davide Bacciu, Federico Errica, Marco Podda, Alessio Micheli
June 2020 In Neural Networks
A Gentle Introduction to Deep Learning for Graphs

A top-down approach to the field of machine learning for graphs.
PDF Cite Code DOI
Federico Errica, Marco Podda, Davide Bacciu, Alessio Micheli
April 2020 In ICLR-2020
A Fair Comparison of Graph Neural Networks for Graph Classification

A fair and robust evaluation of Deep Graph Networks for graph classification tasks.
PDF Cite Code
Davide Bacciu, Federico Errica, Alessio Micheli
July 2018 In ICML-2018
Contextual Graph Markov Model: a Deep and Generative Approach to Graph Processing

A deep, probabilistic, and scalable machine learning model for graphs. Work done as part of my master thesis.
PDF Cite Code

Publications

Quickly discover relevant content by filtering publications.
Antonio Carta, Andrea Cossu, Federico Errica, Davide Bacciu (2022). Catastrophic Forgetting in Deep Graph Networks: A Graph Classification Benchmark. In Frontiers in Artificial Intelligence.

PDF Cite Code DOI

Federico Errica, Giacomo Iadarola, Fabio Martinelli, Francesco Mercaldo, Alessio Micheli (2021). Robust Malware Classification via Deep Graph Networks on Call Graph Topologies. ESANN-2021.

Cite Code

Federico Errica, Fabrizio Silvestri, Bora Edizel, Ludovic Denoyer, Fabio Petroni, Vassilis Plachouras, Sebastian Riedel (2021). Concept Matching for Low-Resource Classification. IJCNN-2021.

PDF Cite Code DOI

Daniele Atzeni, Davide Bacciu, Federico Errica, Alessio Micheli (2021). Modeling Edge Features with Deep Bayesian Graph Networks. IJCNN-2021.

PDF Cite Code DOI

Federico Errica, Davide Bacciu, Alessio Micheli (2020). Theoretically Expressive and Edge-aware Graph Learning. ESANN-2020.

PDF Cite

See all publications

Contact