Computer Science

Photo of Dr Fabrizio Costa

Dr Fabrizio Costa

Senior Lecturer in Data Analytics

 F.Costa@exeter.ac.uk

 (Streatham) 4602 or (Streatham - anastasios rousssos) 4606

 01392 724602 or (anastasios rousssos) 01392 724606

Overview

Fabrizio received his PhD in Computer Science from the University of Firenze, Italy. Before joining Exeter University as a Lecturer in Data Analytics (2017), he worked as a Research Assistant at the Universita' degli Studi di Firenze, Italy, at the Katholieke Universiteit Leuven, Belgium and at the Albert Ludwig University of Freiburg, Germany.

Reserach interests:

At high level my main research question is: how can we integrate domain expert knowledge at the right abstraction level into machine learning (ML) systems and how can we extract knowledge from these systems at various abstraction levels.

In particular I have been interested in: 1) generalizing the type of information that ML algorithms can process and 2) extending the type of problems that ML can address.

1) The study of input generalization for ML algorithms has lead to the development of efficient and flexible graph kernels and feature generators for graphs (see the EDeN library) with applications in relational learning and chemo- and bioinformatics domains.

2) The study of generalizing ML tasks has lead to the notion of data driven design or Constructive Machine Learning (CML) (see the CML workshops). The objective of CML is the design of algorithms that can build novel objects endowed with desired properties. Such algorithms have a wide spectrum of applications, ranging from drug and protein design to automatic software generation, from music composition to video game assets construction.

Academic profiles:

Google Scholar,  DBLP, Research Gate

Back to top

Publications

Copyright Notice: Any articles made available for download are for personal use only. Any other use requires prior permission of the author and the copyright holder.

| 2023 | 2020 | 2019 | 2018 | 2017 | 2016 | 2015 | 2014 | 2013 | 2012 | 2011 | 2010 | 2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 2001 | 2000 | 1999 |

2023

2020

2019

2018

2017

  • Backofen R, Costa F, Theis F, Marr C, Preusse M, Becker C, Saunders S, Palme K, Dovzhenko O. (2017) MicroRNA as an Integral Part of Cell Communication: Regularized Target Prediction and Network Prediction, Information- and Communication Theory in Molecular Biology, Springer.
  • Tran-Van D, Sperduti A, Costa F. (2017) Joint Neighborhood Subgraphs Link Prediction, International Conference On Neural Information Processing, Guangzhou, China, 14th - 17th Nov 2017, Springer.
  • Tran-Van D, Sperduti A, Costa F. (2017) Link Enrichment for Diffusion-based Graph Node Kernels, International Conference on Artificial Neural Networks, Italy, 11th - 15th Sep 2017.
  • Navarin N, Costa F. (2017) An efficient graph kernel method for non-coding RNA functional prediction, Bioinformatics, volume 33, no. 17, pages 2642-2650, DOI:10.1093/bioinformatics/btx295. [PDF]
  • Miladi M, Junge A, Costa F, Seemann SE, Havgaard JH, Gorodkin J, Backofen R. (2017) RNAscClust: clustering RNA sequences using structure conservation and graph based motifs, Bioinformatics, volume 33, no. 14, pages 2089-2096, DOI:10.1093/bioinformatics/btx114. [PDF]
  • Donini M, Navarin N, Lauriola I, Aiolli F, Costa F. (2017) Fast hyperparameter selection for graph kernels via subsampling and multiple kernel learning, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Bruges (belgium), 26th - 28th Apr 2017.
  • Van Dinh T, Sperduti A, Costa F. (2017) The Conjunctive Disjunctive Node Kernel, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN), Bruges (belgium), 26th - 28th Apr 2017.
  • Costa F. (2017) Learning an efficient constructive sampler for graphs, Artificial Intelligence, volume 244, pages 217-238, DOI:10.1016/j.artint.2016.01.006.

2016

2015

  • Alkhnbashi O, Shah S, Garrett R, Saunders S, Costa F, Backofen R. (2015) Towards an automatic annotation of CRISPR-cas subtypes, Annual international CRISPR meeting.
  • Frasconi P, Costa F, De Raedt L, De Grave K. (2015) KLog: A language for logical and relational learning with kernels, IJCAI International Joint Conference on Artificial Intelligence, volume 2015-January, pages 4183-4187.
  • Makarova KS, Wolf YI, Alkhnbashi OS, Costa F, Shah SA, Saunders SJ, Barrangou R, Brouns SJJ, Charpentier E, Haft DH. (2015) An updated evolutionary classification of CRISPR-Cas systems, NATURE REVIEWS MICROBIOLOGY, volume 13, no. 11, pages 722-736, DOI:10.1038/nrmicro3569. [PDF]

2014

2013

2012

  • Frasconi P, Costa F, De Raedt L, De Grave K. (2012) kLog: A Language for Logical and Relational Learning with Kernels, ArXiv.
  • Antanas L, Frasconi P, Costa F, Tuytelaars T, De Raedt L. (2012) A Relational Kernel-based Framework for Hierarchical Image Understanding, Joint IAPR International Workshops on Structural and Syntactic Pattern Recognition (SSPR 2012) and Statistical Techniques in Pattern Recognition (SPR 2012).
  • Heyne S, Costa F, Rose D, Backofen R. (2012) GraphClust: alignment-free structural clustering of local RNA secondary structures, BIOINFORMATICS, volume 28, no. 12, pages I224-I232, DOI:10.1093/bioinformatics/bts224. [PDF]

2011

  • Ramon J, Costa F, Florêncio CC, Kok J. (2011) Preface, Fundamenta Informaticae, volume 113, no. 2, DOI:10.3233/FI-2011-601.
  • Maticzka D, Backofen R, Costa F. (2011) Prediction of binding preferences of RNA-binding proteins using a novel graph representation, 19th Annual International Conference on Intelligent Systems for Molecular Biology and 10th European Conference on Computational Biology (ISMB-ECCB)(poster).
  • Kundu K, Costa F, Sinha R, Reth M, Huber M, Backofen R. (2011) Identification of SH2-peptide interactions using support vector machine, 19th Annual International Conference on Intelligent Systems for Molecular Biology and 10th European Conference on Computational Biology (ISMB-ECCB) (poster).
  • Vens C, Costa F. (2011) Random forest based feature induction, Proceedings - IEEE International Conference on Data Mining, ICDM, pages 744-753, DOI:10.1109/ICDM.2011.121.

2010

  • Ramon J, Costa F, Florencio CC, Kok J. (2010) StReBio'09, ACM SIGKDD Explorations Newsletter, volume 11, no. 2, pages 88-89, DOI:10.1145/1809400.1809418.
  • Kimmig A, Costa F. (2010) Link and node prediction in metabolic networks with probabilistic logic, Analysis of Complex Networks, European Conference on Machine Learning (ECML).
  • Costa F. (2010) Augmented molecular graph kernel QSARs, 5th Joint Sheffield Conference on Chemoinformatics, (poster).
  • Kimmig A, Costa F. (2010) Link and Node Prediction in Metabolic Network with Probabilistic Logic, 7th International Symposium on Networks in Bioinformatics (ISNB).
  • Costa F, De Grave K. (2010) Fast neighborhood subgraph pairwise distance kernel, Proceedings of the 26th International Conference on Machine Learning.
  • Grave KD, Costa F. (2010) Molecular Graph Augmentation with Rings and Functional Groups, Journal of Chemical Information and Modeling, volume 50, no. 9, pages 1660-1668, DOI:10.1021/ci9005035. [PDF]
  • Schietgat L, Costa F, Ramon J, De Raedt L. (2010) Effective feature construction by maximum common subgraph sampling, Machine Learning, volume 83, no. 2, pages 137-161, DOI:10.1007/s10994-010-5193-8. [PDF]

2009

  • Costa Florêncio C, Costa F, Ramon J, Kok J. (2009) Proceedings of the KDD-09 Workshop on Statistical and Relational Learning in Bioinformatics, StReBio '09: Preface, Proceedings of the KDD-09 Workshop on Statistical and Relational Learning in Bioinformatics, StReBio '09.
  • Schietgat L, Costa F, Ramon J, De Raedt L. (2009) Maximum Common Subgraph Mining: A Fast and Effective Approach towards Feature Generation, 7th International Workshop on Mining and Learning with Graphs (MLG).
  • Ramon J, Costa F. (2009) Handling missing values and censored data in PCA of pharmacological matrices, KDD Workshop on Statistical and Relational Learning in Bioinformatics.

2008

2007

2006

  • Costa F, Menchetti S, Ceroni A, Passerini A, Frasconi P. (2006) Decomposition Kernels for Natural Language Processing, Workshop on Learning Structured Information in Natural Language Applications at the 11th Conference of the European Chapter of the Association for Computational Linguistics (EACL).
  • Costa F, Passerini A, Frasconi P. (2006) Learning Structured Outputs via Kernel Dependency Estimation and Stochastic Grammars, Mining and Learning with Graphs (MLG) held at European Conference on Machine Learning (ECML).

2005

  • Menchetti S, Costa F, Frasconi P. (2005) Weighted Decomposition Kernels for Protein Subcellular Localization, Annual Meeting of the Bioinformatics Italian Society (BITS).
  • Cerbioni K, Palanca E, Starita A, Costa F, Frasconi P. (2005) A Grid Focused Community Crawling Architecture for Medical Information Retrieval Services, 2nd International Conference on Computational Intelligence in Medicine and Healthcare, BIOPATTERN Conference.
  • Menchetti S, Costa F, Frasconi P, Pontil M. (2005) Wide coverage natural language processing using kernel methods and neural networks for structured data, Pattern Recognition Letters, volume 26, no. 12, pages 1896-1906, DOI:10.1016/j.patrec.2005.03.011.
  • Menchetti S, Costa F, Frasconi P. (2005) Weighted decomposition kernels, ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning, pages 585-592, DOI:10.1145/1102351.1102425.
  • Costa F, Frasconi P, Lombardo V, Sturt P, Soda G. (2005) Ambiguity resolution analysis in incremental parsing of natural language, IEEE Trans Neural Netw, volume 16, no. 4, pages 959-971, DOI:10.1109/TNN.2005.849837. [PDF]

2004

  • Costa F, Frasconi P. (2004) Distributed community crawling, Proceedings of the 13th International World Wide Web Conference on Alternate Track, Papers and Posters, WWW Alt. 2004, pages 362-363, DOI:10.1145/1013367.1013476.
  • Costa F, Frasconi P. (2004) Distributed community crawling, Thirteenth International World Wide Web Conference Proceedings, WWW2004, pages 1094-1095.

2003

2002

  • Costa F, Frasconi P, Sturt P, Lombardo V. (2002) Exploring the effect of experience on a recursive neural network model of structural preferences, 15th Annual Conference on Human Sentence Processing (CUNY).
  • Costa F, Frasconi P, Lombardo V, Sturt P, Soda G. (2002) Enhancing first-pass attachment prediction, 15th European Conference on Artificial Intelligence (ECAI).

2001

  • Costa F, Sturt P, Lombardo V, Frasconi P. (2001) The effect of experience on a hybrid model of human first-pass attachment preferences, Annual Conference on Architectures and Mechanisms for Language Processing (AMLAP).
  • Sturt P, Lombardo V, Costa F, Frasconi P. (2001) A wide-coverage model of first-pass structural preferences in human parsing, 14th Annual Conference on Human Sentence Processing (CUNY).
  • Costa F, Lombardo V, Frasconi P, Soda G. (2001) Wide coverage incremental parsing by learning attachment preferences, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), volume 2175, pages 297-307.

2000

  • Costa F, Frasconi P, LOmbardo V, Soda G. (2000) Learning to Rank Structured Alternatives: An Application to Incremental Processing of Natural Language, Workshop on Attribute-Value and Relational Learning: Crossing the Boundaries (ICML).
  • Costa F, Frasconi P, Lombardo V, Soda G. (2000) Learning to ranking structured alternatives: an application to incremental processing of Natural Language, Conference of the Italian Association for Artificial Intelligence (AIIA).
  • Costa F, Frasconi P, Lombardo V, Soda G. (2000) Learning Incremental Syntactic Structures with Recursive Neural Networks, Conference on Knowledge-Based Engineering Systems (KES), volume 2, pages 458-461.
  • Sturt P, Lombardo V, Costa F, Frasconi P. (2000) An Experience-Based Model of Incremental Parsing using Dynamic Grammars and Recursive Neural Networks, Annual Conference on Architectures and Mechanisms for Language Processing (AMLAP).

1999

  • Costa F, Frasconi P, Soda G. (1999) Recursive Connectionist Networks for Trees with Large Outdegree, Conference of the Italian Association for Artificial Intelligence (AIIA).
  • Costa F, Frasconi P, Soda G. (1999) A topological transformation for hidden recursive models architecture networks, ESANN.

Back to top

Further information

Project proposals

I am interested in supervising projects related to machine learning in structured domains (i.e. domains where data is best encoded as strings, trees or graphs) and in particular in the problem of constructive machine learning or machine learnable design tasks. Examples of such problems range from learning how to design molecular compounds to learning how to design web pages based primarily on samples of good and bad solutions.  

In general I am interested in supervising projects that have to do with constructing or generating structured instances, using a mixture of combinatorial techniques and machine learning, so if you are interested in a specific applicaiton domain let me know.

Here are some project ideas:

[1] Learning to construct computer game assets.

Characteristic elements of a computer game such as maps, characters, weapons, entire level descriptions and even music can be designed by human experts or generated automatically using procedural content generation (PCG) techniques. These generative methods are often quite constrained via hand-coded heuristics to ensure high quality.

In this project the candidate will develop strategies to encode computer game assets as graphical structures and develop predictive techniques under the Bayesian optimization framework to automatically improve their design based on user responses.

 

[2] Learning to personalize user curricula.

When students want to learn a novel concept, access to encyclopedic knowledge, even if organized as a network of linked articles (e.g. Wikipedia), is not enough: students might not know how to autonomously navigate a vast body of knowledge, and if they do not perceive that they are moving toward their goal of understanding at a sufficiently fast pace they become demotivated. Traditionally one would use a course book which orders topics and concepts in such a way as to allow to grasp subsequent concepts given that the preceding ones have been understood. However, a course book is a one-size-fit-all solution and does not take into account a student’s prior knowledge. Increased efficiency in learning could be obtained by personalizing the curriculum taking into account the student's background knowledge and skills. An optimal sequence of concepts to learn can be obtained by identifying at each step the concepts that lie on the surface that separates understood concepts from non understood concepts ('borderline' concepts). In Machine Learning a popular strategy for 'active learning' is to acquire supervised information on the elements closest to the current decision surface of the learning algorithm. In some sense, these elements are the most informative.

In this project the candidate will formulate the identification of the borderline concepts as an optimization problem over a network of available concepts.

[3] Learning to optimize a web page design.

One way to simplify the construction of web pages is using some form of sketch that can later be transformed in fully functional code. Currently, the translation from web pages mockups to code is  mainly done via rigid parametric systems such as e.g. SquareSpace or WebFlow. Recently there have been attempts to apply machine learning techniques to map web pages mockups directly to code (AirBnB's Sketching Interfaces or Beltramelli's pix2code) using deep neural networks systems. These end-to-end approaches however, while more flexible, require considerable post-processing effort to obtain viable code. An alternative approach is to ask the user to rank alternative web designs in order of preference and automatically generate a solution in an extremely large combinatorial space with stronger guarantees of correctness.

In this project the candidate will develop strategies to represent the code underlying web pages as graphical structures, use flexible graph grammars to automatically capture the syntactic constraints and use machine learning systems based on graph kernel to learn user preferences to generate improved versions.  

[4] Interactive image generation

It is easy for humans to rank images according to a personal sense of aestetics (i.e. how much one likes specific images). It is however more difficult to explicitly state the reasons for such choices. Machine Learning techniques can be used to automatically extract what are the elements that characterize a set of preferred items and generative systems can be employed to build images that use preferentially the identified elements in novel combinations. In this project a user will start from a collection of selected images and will be able to interact with an image generator capable to build novel but tailored images. The system will consider the user feedback as to which part of the image needs to be changed and will allow the continous generation of complex but pleasing images using a simple point and click interface.

In this project the candidate will 1) develop segmentation techniques to automatically identify relevant elements in images, 2) convert images to networks of connected elements, 3) employ graph grammars (F. Costa, Learning an efficient constructive sampler for graphs. Artificial Intelligence 2017) and graph kernel machine learning techniques (F. Costa, K. De Grave. "Fast neighborhood subgraph pairwise distance kernel.” ICML 2010) to induce the generative rules and inform an optimization algorithm on which actions to perform to generate a novel image. 

Back to top