Ofertas de proyectos

A list of issues hard to compute would include:

1. The computation of the partition function is NP-hard, involving an  exponential sum of terms
2. The exact computation of the derivative of the log-likelihood is  also NP-hard, since it contains the derivative of the partition function

Therefore, in practice we have to approximate both the computation of
the probabilities and several components of the learning process
itself. These drawbacks have prevented RBMs to show their real
potential as truly probabilistic models.

Currently, we are working on trying to improve several of the unsolved
issues related to RBMs:

1. - Mechanisms to control the learning process: www.lsi.upc.edu/~eromero/Publications/Downloads/2018-tnnls-stopcritRBM.pdf
2. Better approximations of the derivative of the log-likelihood: http://www.lsi.upc.edu/~eromero/Publications/Downloads/2019-nn-weightedCD.pdf
3. Efficient approximation of the partition function (work in progress)

These works have opened new lines of research, some of which can be
the topic of a Master's Thesis. The scope and degree of depth of the
work can be adapted to the estimated times to complete the Thesis. For
further details, contact Enrique Romero (eromero@cs.upc.edu).

This is precisely why companies such as Google have invested a lot of resources on exploring applications of GNN, some of which are as mediatic as Alpha Fold (recently solving the Protein-Folding problem), or used by Goolge Maps to make predictions of the expected traffic of a road. In this regard, the goal of this thesis is to continue investigating about the potential of GNNs when applied to the field of Computer Networks. Our goal is to provide GNN-based solutions that make computer networks run more autonomously and in a more efficient way.

 The goal of this thesis is to enable the application of Deep Reinforcement Learning techniques for optimization problems in very large and complex scenarios, in this case Data Center Networks. Inspired by the last related work from OpenAI (https://openai.com/blog/evolution-strategies/), we will explore time-efficient alternatives to train a Graph Neural Network based DRL agent on large optimization scenarios in a reasonable amount of time.

A more detailed description of the project can be found in

https://www.cs.upc.edu/~eromero/Downloads/Retina-TFM-Project-01.pdf

The project is proposed in collaboration with Javier Zarranz Ventura
(Institut Clínic d'Oftalmologia, ICOF, Hospital Clínic de Barcelona, and
Institut d'Investigacions Biomèdiques August Pi I Sunyer, IDIBAPS),
which would provide a large annotated database to develop the project. For further information, please contact Alfredo Vellido (avellido@cs.
upc.edu) or Enrique Romero (eromero@cs.upc.edu).

Recently, diffusion-based image generative models have obtained amazing results in multiple tasks such as text-conditioned image generation. They have been also successfully applied to image translation tasks, in which an input image is converted into a target domain defined by, e.g., the appearance of an example image (style transfer), text condition or in other ways. Image cartoonization is a particular case of image translation in which the goal is to apply to the input image a cartoon style. While full image or video cartoonization could be seen as a threat by animation professionals, the possibility to use cartoonized image or video backgrounds (without people) would be an innovation opportunity.

Development of an automatic plan generation alternative to

  https://arxiv.org/abs/2004.13204v1

  http://staff.ustc.edu.cn/~fuxm/projects/DeepLayout/index.html

with Reinforcement Learning

The syntactic structure of a sentence can be represented as a tree where vertices are words and arcs indicate syntactic dependencies between words (Kübler et al 2012). In languages, these structures are typically planar, namely no two arcs cross when drawn above the sentence (Ferrer-i-Cancho et al 2018). Another property of these structures is that the distance between syntactically related words is much smaller than expected in a shuffling the sentence (Ferrer-i-Cancho et al 2022), a fact that lead to the formulation of the syntactic dependency distance minimization principle (Ferrer-i-Cancho 2004).

Syntactic dependency parsing is the branch of computational linguistic concerned with the extraction of syntactic dependency structures from raw text (Kübler et al 2012). This research proposal is focused on unsupervised syntactic dependency parsing, namely methods to extract syntactic dependency structures from unlabelled data (Han et al 2020). Although these methods can be extremely complex, here we aim to start from simple by taking Yuret's (1998) proposal as starting point to build from it more complex methods. Yuret's (1998) unsupervised parser consists of three steps:
1) training, i.e. learning the statistics of word co-occurrences in texts
2) generating a weighted graph where all vertices are the words that have appeared during training and weights are the mutual information between word pairs.
3) parsing a new sentence and thus obtaining its syntactic dependency structure. This structure is the maximum spanning tree from the subgraph of the weighted graph in Step 2 that is induced by the words of the sentence.

In Yuret's proposal, the maximum spanning tree in Step 3 is restricted to be planar. Interestingly, it has been argued that planarity may be a side-effect of pressure to reduce the distance between syntactically related word rather than a direct pressure of sentence to be planar (Gómez-Rodríguez & Ferrer-i-Cancho 2017, Gómez-Rodriguez, Christiansen & Ferrer-i-Cancho 2022). Therefore, the main goal of this project is to explore alternative unsupervised parsing methods:
a) an unsupervised parser that does not impose planarity. That is, Step 3 consists of computing a maximum spanning tree using classic algorithms for calculating maximum spanning trees.
b) an unsupervised parser that does not impose planarity but incorporates syntactic dependency distance minimization and thus tends to produce planar syntactic dependency structures naturally.

The bulk of the project consists of implementing simple parsers of the sort of a-b) and evaluating their performance. The parsers will be applied to human languages and also to sequences that other species produce (great apes and cetaceans).

This project will be developed in the environment of the LQMC research group https://lqmc.upc.edu/.

REFERENCES

Ferrer-i-Cancho, R (2004). Euclidean distance between syntactically linked words. Physical Review E 70, 056135.
Gómez-Rodríguez, C. & Ferrer-i-Cancho, R. (2017). Scarcity of crossing dependencies: a direct outcome of a specific constraint? Physical Review E 96, 062304.
Gómez-Rodríguez, C., Christiansen, M. & Ferrer-i-Cancho, R. (2022). Memory limitations are hidden in grammar. Glottometrics 52, 39-64.
Ferrer-i-Cancho, R. Gómez-Rodríguez, R., Esteban, J.L. (2018). Are crossing dependencies really scarce? Physica A: Statistical Mechanics and its Applications 493, 311-329.
Ferrer-i-Cancho, R., Gómez-Rodríguez & Esteban, J.L. & Alemany-Puig, L. (2022). Optimality of syntactic dependency distances. Physical Review E 105, 014308.
Han, W., Jiang, Y., Tou Ng, H. & Tu, K. (2020). A survey of unsupervised dependency parsing. In Proceedings of the 28th International Conference on Computational Linguistics, pages 2522¿2533, Barcelona, Spain. International Committee on Computational Linguistics.
Kübler, S., McDonald, R. & Nivre, J. (2009). Dependency parsing. Synthesis Lectures on Human Language Technologies 1, 1-127.
Yuret, D. (1998). Discovery of linguistic relations using lexical attraction. PhD Thesis, MIT.

The application of Artificial Intelligence (AI) and Machine Learning (ML) to network security (AI4SEC) is paramount against cybercrime. While AI/ML is mainstream in domains such as computer vision and natural language processing, traditional AI/ML has produced below-par results in AI4SEC. Solutions do not properly generalize, are ineffective in real deployments, and are vulnerable to adversarial attacks. A fundamental limitation is the lack of AI/ML technology specific to network security. Due to their unique ability to learn and generalize over graph-structured information, graph-learning approaches, and in particular Graph Neural Networks (GNNs), have recently enabled groundbreaking applications in multiple fields where data are generally represented as graphs. Network security data are intrinsically relational, and initial research suggests that graph-structured representations and GNNs have the potential to become foundational to AI4SEC, in the way convolutional and recursive networks were to computer vision and natural language processing.
The goal of GRAPHS4SEC is to leverage graph data representations and modern GNN technology to conceive a new breed of robust GNN-based network security methods which could radically advance the AI4SEC practice. The objectives of GRAPHS4SEC are: (a) to investigate algorithmic methods that facilitate modeling and learning from graph-based network security data; (b) to compare the benefits and overheads of GNN-based AI4SEC to traditional AI/ML in terms of detection performance, generalization, scalability, and robustness against adversarial attacks; (c) to showcase the benefits and improvements of GRAPHS4SEC technology in four critical, real-world network security applications with significant impact for society, considering (in particular) the detection and early mitigation of phishing and fake/malicious websites, a threat among the most popular and society-wide harmful in today's Internet.

LABINQUIRY project aims to develop a didactic infrastructure to help secondary school and university teachers to implement inquiry-based learning processes called study and research paths (SRPs). The starting point of an SRP is an open question that students address in small groups with the help of the teacher, they have to search for data, information, new tools and knowledge, etc, study them and jointly elaborate a single answer of the whole class.

One difficulty posed by the SRPs in terms of management is the teacher's treatment of the weekly reports produced by the student groups with the partial results of the work done during the week. The teacher responsible for the SRP has to "process" these reports in limited time in order to prepare the next class session. In this processing, the teacher has to read the reports - about 10-12 per class -, compile the most important information and group them according to the type of questions raised by each team, the data collected, the interim results provided, and the sources of information used.

The research group responsible for the LABINQUIRY project is interested in achieving an automatic processing of the content of student reports. The aim of the TFM would be, from a set of 10-12 reports, to collect and organise the information contained in the reports to help the teacher organise the next task. Both the questions posed by the teacher in the SRP and the students' reports are natural language texts. The collection and organisation of the information can be done at different levels and from different points of view (some of which are already fixed but some of which are not), giving flexibility to the work to be done during the TFM. In principle, the techniques to be used would fall into the following areas: unsupervised learning (clustering) and natural language processing (word embeddings).

The expected timing is January-June 2024.

To carry out this project it is advisable to have a good mastery of Spanish and/or Catalan and the techniques mentioned above.