Giorgio Vinciguerra

Giorgio Vinciguerra

PhD candidate

Università di Pisa

Biography

I’m a PhD candidate in Computer Science at the University of Pisa, currently a member of the A³ Lab led by Prof. Paolo Ferragina.

I’m working on “learning-based compressed data structures”, that is, data structures that achieve new space-time trade-offs compared to traditional solutions by learning, in a rigorous and efficient algorithmic way, the regularities in the input data with tools from machine learning and computational geometry.

My research falls under a national project named Multicriteria data structures funded by the Italian Ministry of University and Research and under the European H2020 project named SoBigData++.

Interests
  • Compact data structures
  • Algorithm engineering
  • Data compression
Education

  • M.Sc. in Computer Science, 2018

    University of Pisa

  • B.Sc. in Computer Science, 2016

    University of Pisa

Papers

Paolo Ferragina, Giovanni Manzini, Giorgio Vinciguerra. Repetition- and linearity-aware rank/select dictionaries. ISAAC, 2021.
Cite Code DOI
Paolo Ferragina, Fabrizio Lillo, Giorgio Vinciguerra. On the performance of learned data structures. Theor. Comput. Sci., 2021.
PDF Cite Code DOI
Paolo Ferragina, Giorgio Vinciguerra. Learning Based Compressed Data Structures. Stanford Compression Workshop, 2021.
Poster Abstract PDF
Antonio Boffa, Paolo Ferragina, Giorgio Vinciguerra. A “learned” approach to quicken and compress rank/select dictionaries. ALENEX, 2021.
PDF Cite Code DOI Experiments code & datasets
Paolo Ferragina, Fabrizio Lillo, Giorgio Vinciguerra. Why are learned indexes so effective?. ICML, 2020.
PDF Cite Code Slides Video
Paolo Ferragina, Giorgio Vinciguerra. The PGM-index: a fully-dynamic compressed learned index with provable worst-case bounds. PVLDB, 2020.
PDF Cite Code Project Slides Video DOI
Paolo Ferragina, Giorgio Vinciguerra. Learned data structures. Recent Trends in Learning From Data (Springer), 2020.
PDF Cite DOI

Projects

Block-ε tree

Block-ε tree

A compressed rank/select dictionary exploiting approximate linearity and repetitiveness.

LA-vector

LA-vector

A compressed bitvector/container supporting efficient random access and rank queries.

PyGM

PyGM

Python library of sorted containers with state-of-the-art query performance and compressed memory usage.

PGM-index

PGM-index

A data structure enabling fast searches in arrays of billions of items using orders of magnitude less space than traditional indexes.

CSS-tree

CSS-tree

A C++11 implementation of the Cache Sensitive Search tree.

NN Weaver

NN Weaver

A Python library to build and train feedforward neural networks, with hyperparameters tuning capabilities.

Talks

The design of learning-based compressed data structures
16 Aug 2021 14:30 (UTC+02) — 14:50 (UTC+02) Copenhagen, Denmark (Video conference)
Slides
A tutorial on learning-based compressed data structures
12 Apr 2021 08:00 (UTC+02) — 09:00 (UTC+02) University of Melbourne (Video conference)
Theory and practice of learning-based compressed data structures
19 Mar 2021 10:00 (UTC+01) — 11:00 (UTC+01) Université de Lille (Video conference)
Slides
Learned indexes
21 Jan 2020 16:00 (UTC-04) — 16:45 (UTC+04) DASlab @ Harvard SEAS
The PGM-index: a multicriteria, compressed and learned approach to data indexing
6 Sep 2019 11:00 (UTC+02) — 11:45 (UTC+02) Schloss Dagstuhl – Leibniz-Zentrum für Informatik
See all events

Teaching & Supervision

Teaching assistant for:

I co-supervised:

  • Antonio Boffa, Spreading the learned approach to succinct data structures, MSc in Computer Science - ICT, 2020.
  • Alessio Russo, Learned index per i db del futuro, BSc in Computer Science, 2020.
  • Lorenzo De Santis, On non-linear approaches for piecewise geometric model, MSc in Computer Science - AI, 2019.

Knowledge is like a sphere; the greater its volume, the larger its contact with the unknown.

― Blaise Pascal

Contact