Ofertes de projectes

MAI

ENRIQUE ROMERO MERINO

CS

Boltzmann Machines are probabilistic models developed in 1985 by
D.H. Ackley, G.E. Hinton and T.J. Sejnowski. In 2006, Restricted
Boltzmann Machines (RBMs) were used in the pre-training step of
several successful deep learning models, leading to a new renaissance
of neural networks and artificial intelligence.

In spite of their nice mathematical formulation, there are a number of
issues that are hard to compute:

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).

Good programming skills.

MAI

ALFREDO VELLIDO ALCACENA y ENRIQUE ROMERO MERINO

CS

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).

MAI

PETIA RADEVA (Universitat de Barcelona)

UB

The traits variation can be studied in the population by genotyping individuals, which helps us determine the DNA sequence at specific positions within the genome and identify small variations called single nucleotide polymorphisms (SNPs). Large-scale efforts to genotype individuals have been made thanks to advancements in sequencing technology over the past decade. As a result, we now have several large genome-wide association studies (GWAS) available that include both SNP and phenotypic data from thousands of individuals. These datasets have already been instrumental to help scientists identify genes associated with several diseases and

to develop novel therapeutics.

A set of SNPs can be naturally expressed as a graph with the SNPs representing the nodes and the edges representing interactions between them (e.g., at the protein level). Thus, it stands to reason that the application of graph neural networks (GNNs) [SGT+08] to perform phenotypical prediction leveraging a SNP graph could be a sound strategy. In fact, recent developments in GNNs have increased their capabilities and expressive power, finding applications in a number of areas such as antibacterial discovery [SYS+20] or fake news detection [MFE+19]. In contrast to most state of the art methods, graph neural networks can perform many tasks at the same

time, including node prediction (e.g., SNP class), edge prediction (SNP-SNP interactions), and graph prediction (phenotype). Nevertheless, the application of GNNs to this problem remains largely unexplored. Thus, gaining understanding of their performance in this context could lead to critical insights for the application of GNNs in other settings, as well as novel biological discoveries with potential implications in disease.

Goals

1. Review state of the art methods for phenotype prediction.

2. Develop a graph neural network algorithm to perform phenotype prediction and benchmark against the state of the art using both simulated and real GWAS data. Students should take advantage of available python frameworks and packages like Pytorch and networkx.

3. Study potential attribution techniques to assign ranked importance values to parts of a graph that are relevant in determining the phenotype.

References

[MFE+19] Federico Monti, Fabrizio Frasca, Davide Eynard, Damon Mannion, and Michael M Bronstein. Fake news detection on social media using geometric deep learning. arXiv preprint arXiv:1902.06673, 2019.

[SGT+08] Franco Scarselli, Marco Gori, Ah Chung Tsoi, Markus Hagenbuchner, and Gabriele Monfardini. The graph neural network model. IEEE transactions on neural networks, 20(1):61¿80, 2008.

[SYS+20] Jonathan M Stokes, Kevin Yang, Kyle Swanson,Wengong Jin, Andres Cubillos-Ruiz, Nina M Donghia, Craig R MacNair, Shawn French, Lindsey A Carfrae, Zohar Bloom-Ackermann, et al. A deep learning approach to antibiotic discovery. Cell, 180(4):688¿702, 2020.

Requisits: Machine learning.
Additionally, Deep learning is a merit.

This project ideally should be developed by a team of three people. In case the team is of less members the project will be properly rescaled considering only some of the goals.

MAI

MIQUEL SANCHEZ MARRE

CS

This combination of processes may include waste material separations (to recover useful materials) but also more complex chemical transformations of waste to generate other materials, that can be used as resource materials in other processes. This combination of possible processes and materials is really high and it is very difficult to assess which are the best possible solutions regarding the minimization of economic costs and minimizing the ecological print. Optimal transformation will consider several characteristics like: transformation cost, recycled product purity, transformation process efficiency, transport cost, distance between companies, etc.

 

The main functionality of the intelligent planner is to recommend the best path (series of transforming steps) to be followed to transform a residual product A into a recycled raw material B. In addition to that, the system will include the functionalities of adding/removing/modifying residual products, recycled products and/or transformation processes to the system ontology. The system ontology will be based on the use, exploitation and extension of the OntoCAPE ontology[1]. Thus, a good understanding of the OntoCAPE ontology will be necessary to be able to manipulate it for extracting knowledge, using reasoning mechanisms, and possible, being able to add new information/knowledge to the ontology.

 

This work will be deployed within the framework of a Spanish Research project entitled "Circular Economy in the Process Industries: Methods and Tools for Automatic Recommendation of Process Technologies enabling Circular Integration (CEPI) [PID2020-116051RB-100]" in collaboration between the Intelligent Data Science and Artificial Intelligence Research Centre (IDEAI-UPC) with Centre for Process and Environment Engineering Research Group (CEPIMA) at UPC.

 

Additional information:

• https://en.wikipedia.org/wiki/OntoCAPE
• https://www.avt.rwth-aachen.de/cms/AVT/Wirtschaft/SoftwareSimulation/~ipts/OntoCape
• https://www.springer.com/gp/book/9783642046544
• https://www.sciencedirect.com/science/article/abs/pii/S0952197606001199

Load work: This work will have a workload-equivalent of 3 months (450h) to generate a prototype. This work is the equivalent to the workload of a Final Project of a Master/Master's Thesis (TFM). Notwithstanding, the work can be spread along the year 2022 and 2023 until when the TFM must be presented.

 

Advisor/s of the work: the work will be advised by Prof. Antonio Espuña (antonio.espuna@upc.edu) from Dept. of Chemical Engineering and CEPIMA research group (Centre for Process and Environment Engineering, https://cepima.upc.edu/en) and by Dr. Miquel Sànchez-Marrè (miquel@cs.upc.edu), from Dept. of Computer Science and IDEAI (Intelligent Data Science and Artificial Intelligence, https://ideai.upc.edu/en) Research Centre.

 

Funding: There is the possibility that the work would be funded within the framework of the Research project on "Circular Economy in the Process Industries: Methods and Tools for Automatic Recommendation of Process Technologies enabling Circular Integration (CEPI) [PID2020-116051RB-100]" depending on the candidates' dedication load and/or candidates' advances.

 

Contact: Antonio Espuña (antonio.espuna@upc.edu) and

                Miquel Sànchez-Marrè (miquel@cs.upc.edu).

[1] OntoCAPE is a large-scale ontology for the domain of Computer-Aided Process Engineering (CAPE).

Basic Requirements: The students should own a Graduate in Computer Science or equivalent, and they should be enrolled in the Master of Artificial Intelligence (UPC-URV-UB) or equivalent, being able to register and matriculate the TFM for the course 2022/2023.

Other requirements: Good oral and writing skills of English language. Additional Catalan and/or Spanish language skills can be positively evaluated.

MAI

ENRIQUE ROMERO MERINO

CS

Gradient descend based RBM learning is a computationally expensive procedure that requires the computation of a probability normalization constant (called the Partition function) that involves the computation of an exponentially large number of terms. In order to bypass that, different approximations exist, the most celebrated one being the Contrastive Divergence (CD) algorithm [1].

In this project we want to explore an alternative RBM model based on archetype (ARC) learning, described in [2]. The student will have to implement both the standard CD and the ARC algorithms in CUDA, and compare their performance in common benchmarking datasets.

[1] "Training restricted Boltzmann machines: An introduction",
Asja Fischer and Christian Igel, Pattern Recognition 47 (2014) 25-39

[2] "The emergence of a concept in shallow neural networks"
Elena Agliari, Francesco Alemanno, Adriano Barra, Giordano De Marzo, Neural Networks 148 (2022) 232-253

MAI

ALBERTO CABELLOS APARICIO

AC

 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.

- Some knowledge on Machine Learning tools and techniques (e.g, Tensorflow, backpropagation, etc)
- Some knowledge on networking (preferred, but not required)
- Good oral/writing skills
- Applicants please send your CV and academic transcripts to

MAI

PERE BARLET ROS

AC

The main objective of this project is to apply the most recent advances in Deep Learning, including Deep Reinforcement Learning and Graph Neural Networks, to uncover the particular methods used to track Internet users and collect PI.

This project will be useful for both Internet users and the research community, and will produce open source tools, real data sets, and publications revealing most privacy attempting practices.

Some preliminary results of our work in this area were recently published in Proceedings of the IEEE (IF: 9.237) and featured in a Wall Street Journal article.

More info at:

http://personals.ac.upc.edu/pbarlet/papers/web-tracking.survey2015.pdf

http://blogs.wsj.com/digits/2015/08/04/7-ways-youre-being-tracked-online-and-how-to-stop-it/