Ofertes de projectes

Many problems in data science come in the form of binary variables. These may express presence/absence, true/false, yes/no, or be part of higher-order codings. Surprisingly, this case has not attracted mainstream attention in predictive machine learning. However, the existence of specialized learners for this kind of data would be a very welcome addition to data science methods.
The goal of this project is to research the literature and then design new or modified algorithms for binary/categorical datasets  (both for classification and regression problems). The methods will be illustrated experimentally in synthetic and in real-world challenging datasets.

Quantum computers promise exponential improvements over conventional ones due to the extraordinary properties of qubits. However, a quantum computer faces many challenges relative to the movement of qubits which is completely different from the movement of classical data. This thesis delves into these challenges and proposes solutions to create scalable quantum computers enabled by quantum communications, following the current European projects at N3Cat (www.n3cat.upc.edu) on scalable quantum computing.

The interested candidate will work in a group of several PhD students and in collaboration with Universitat Politècnica de València, working in ONE of the following areas:

• Designing classical (bit) and quantum (qubit) communication protocols and network architectures for communications inside a quantum computer.
• Assisting in the modeling of quantum short-range communications within our simulation framework, which is based on NetSquid.
• Developing qubit mapping, partitioning, and routing for scalable quantum computers.
• Study the use of AI to model quantum computers or develop qubit mapping and routing schemes.
• Study the potential adaptation of the existing quantum computing architectures to specific quantum machine learning algorithms (QAOA, VQE, others)

Desenvolupament d'una app web per la integració d'un simulador d'Edge Computing (EC). El funcionament del simulador és el següent: pren com entrada un model que consisteix en nombre de servidors amb uns recursos i un nombre de treballs amb requeriments de recursos els quals han d'estar processats abans d'un temps donat, a més s'introdueix el temps de simulació de dies expressat en minuts. Després de la simulació s'ha de visualitzar les mètriques de jobs completats i els recursos utilitzats. Aquest TFG contempla les següents tasques:

- Determinar quina és l'arquitectura del sistema més adient per la integració.

- Requisits no funcionals:

- Usabilitat.
- Suportabilitat.
- Seguretat.

- Requisits funcionals:

- Accés via Web.
- Gestió d'usuaris.
     Alta/baixa usuaris del simulador.
    Control sessions i d'historial de resultats.

   Usuari administrador desplegament de versions del simulador.
   Control del simulador: carga els models, introduir paràmetres simulador, visualitzar resultats, control simulació.

En primer lloc, s'analitzarà l'estructura de l'eina docent, a partir d'experiències anteriors del grup i de les funcions.

En segon lloc, es plantejaran els diferents menús de l'eina.

En tercer lloc, caldrà implementar les funcionalitats bàsiques. Seria desitjable ja plantejar algun dels problemes i la seva resolució, però dependrà de l'evolució en les anteriors fases.

Es disposa de suport econòmic.

Posar-se en contacte amb manel.mateo@upc.edu per a més informació.

Es vol desenvolupar una aplicació que permeti visualitzar en Realitat Augmentada un model anatòmic (llegint un QR o indicant posició en espai). Es podrà inspeccionar el model anatòmic canviant el punt de vista extern.

El model tindrà un conjunt de punts d'interès (coneguts i indicats sobre el model) que l'usuari podrà seleccionar i que transportaran a l'usuari a una vista interior del put d'interès. Aquesta vista interior es convertirà en una visualització 360 per a què l'usuari pugui mirar al seu voltant.

L'aplicació haurà de permetre a l'usuari iniciar l'experiència, decidir el canvi a 360 i retornar del 360 a la visió de RA.

Virtual reality is becoming a tool for helping patients with several illnesses (strokes, frozen shoulder...) to engage in the rehabilitation exercises.

The purpose of this project is the creation of a serious game that helps patients to engage in rehabilitation exercises for the frozen shoulder.

Missing values have always been a real headache in statistics, machine learning, data mining and now data science. Essentially no one knows what would be the "best" or "right" imputation procedure when confronted with this problem.

To make things worse, sometimes the amount of missing information is so large or evenly distributed that we simply cannot go on with the modelling. In these cases, deletion or removal of some rows/columns containing missing values could be considered as a sensible step prior to imputation. Unfortunately, the order in which this is done greatly affects the quality of the final dataset.

The goal of this project is to design a sensible and efficient algorithm that computes the best deletion procedure for a dataset, taking into account factors as fraction of rows/columns with missing data, importance of variables, total missing percentage, etc.

Computing systems are ubiquitous in our daily life and have transformed the way we learn, work, or communicate with each other, to the point that progress is intimately tied to the improvements brought by new generations of the processors that lie at the heart of these systems. 

A common trait of current computing systems is that their internal data communication has become a fundamental bottleneck. Traditional interconnects are insufficient and threatening to halt progress unless fast and versatile communication alternatives are developed.

In the NaNoNetworking Center in Catalunya (N3Cat, www.n3cat.upc.edu), we envisage a revolution in computer architecture enabled by the use of wireless communications (and other unconventional interconnect technologies) inside computer chips, both classical and quantum. Of course, this vision comes with its own set of challenges since we need to figure out the best way to architect these systems with the new capabilities of the interconnects. In this direction, the candidate will work on one of the following areas:

• Analysis of the communications workload in nowadays computer chips and development of network-on-chip architectures to serve such traffic optimally.
• Development of chiplet-based architectures in the classical domain in cache-coherent based systems, delving into the cache coherence protocol and the interface with the wireless interconnect.
• Development of a supercomputer-in-a-package, a chiplet-based architecture based on message passing, delving into how to speed up the MPI primitives with wireless communications.
• Study of scalable quantum computing systems enabled by wireless classical communication across the chips of the quantum computer and inside a cryogenic refrigerator.

A radial basis function network is an artificial neural network that uses radial basis functions as activation functions. The output of each hidden neuron is an inverse function (e.g. a Gaussian) of Euclidean distances between the inputs and the neuron parameters (thereby acting as prototypes).

Radial basis function networks have many uses, including function approximation, time series prediction, classification, and system control. Compared to MLP networks, they are fast to train, more interpretable, and easier to analyze mathematically.

As many neural networks, there is often the need to introduce a regularization mechanism to avoid overfitting the data. The determination of the regularization parameter (lambda) is typically carried out by trial-and-error, a method that leaves much to be desired.

The work of MJL Orr was a big step towards an automatic way of determining lambda. However, his algorithm still needs some human intervention and lacks mathematical rigour from the point of view of optimization.
The purpose of this project is to progress in the direction of Orr's work and produce a better learning procedure, implement it and test it in real applications. This project is totally research-oriented.

Seeing that not all neural networks fit to all problems, and that relational data is present in a wide variety of aspects of our daily life, the main focus of this thesis in N3Cat (www.n3cat.upc.edu) and BNN-UPC (www.bnn.upc.edu) is to explore the possibilities of the Graph Neural Networks (GNNs), whose aim is to learn and model graph-structured relational data. We are looking for students willing to study the uses, architectures, and algorithms of GNNs. To this end, the candidate will work on ONE of the following areas:

• Uses: Applying GNNs in emerging application areas, including but not limited to (1) electroencephalogram (EEG) analysis for Alzheimer's disease detection, epilepsy classification, motor imagery; (2) acceleration of the compilation of quantum computing algorithms
• Architectures: Investigating ways to accelerate the processing of GNNs in multiple computing platforms (CPU, GPU, accelerators).
• Algorithms: Developing meta-learning data-driven models to estimate the accuracy of a GNN for a given graph, without training, through synthetic graph generation and correlation analysis.

Recent advancements in nanotechnology have enabled the development of means for sensing and wireless communications with unprecedented miniaturization and capabilities, to the point that they can be introduced into the gastrointestinal tract inside a pill or into the bloodstream in the form of passively flowing nanomachines. 

This opens the door to the idea of intra-body communication networks, this is, a swarm of nanosensors inside the human body that use communications to coordinate their actions to sense and localize specific events (lack of oxygen, biomarkers, etc). This can lead to the development of applications such as continuous monitoring of diabetes, detection and localization of cancer micro-tumors, or early detection of blood clots. These possibilities are currently investigated by our team at the N3Cat (www.n3cat.upc.edu).

In this context, We are looking for excellent and self-motivated individuals who are eager to develop ideas that will have a positive societal or socio-economic impact. In particular, the students will work on ONE of the following areas:

• Developing means of zero-power communications between nanosensors inside the human body.
• Designing protocols for communications from nanomachines inside the human body and implants or antennas in the skin surface.
• Studying mobility inside the gastrointestinal tract and human bloodstream towards creating mobility models for accurate simulation of intra-body communications.
• Dimensioning the system (inferring the required nanomachine density) to achieve a particular sensing goal/application.
• Developing AI schemes for the localization of events inside of the human body.

Es vol desenvolupar una aplicació de Realitat augmentada que visualitzi un circuit i permeti conduir un cotxe per aquest circuit. Aquesta aplicació haurà de permetre més d'un jugador, per tant caldrà que un dels jugadors (el primer) es converteixi en servidor i permeti que altres puguin connectar-se a la mateixa carrera (amb cotxes de diferents colors, per exemple). Caldrà també implementar la comunicació necessària entre els diferents participants i el servidor. Cada usuari haurà de disposar de controls per fer moure el seu cotxe. Caldrà que s'inclogui detecció de col·lisions entre els diferents vehicles per simular la carrera.

The Accelerators and Communications for HPC Group leads cutting-edge research and development around accelerators/coprocessors in HPC and high-performance networking. We collaborate closely with the major vendors in the topic for HPC: NVIDIA and Intel. We organize locally international events such as the PUMPS Summer School, PATC Courses on CUDA/OpenACC, and annual hackathons, and collaborate in the organization of related international conferences and workshops such as SC, IEEE Cluster, or AsHES.

The group has been recently awarded one of the prestigious ERC Consolidator Grants. We are looking to cover a PhD Student positions to work on novel system architectural proposals applied to Homomorphically Encrypted Deep Learning Inference. Prospective candidates may take a look at the following paper to get an initial idea of our focus:

https://arxiv.org/abs/2103.16139

Group Web Page: https://www.bsc.es/discover-bsc/organisation/scientific-structure/accelerators-and-communications-hpc

Key Duties

• Drive research in the above-mentioned topic, under the supervision of the Group Manager and the mentorship of a postdoctoral researcher

Cybersecurity is becoming an increasingly important challenge for all companies and individuals alike. While big names used to be the main targets in the past, as people's lives move online, anyone is nowadays a potential target for any kind of cyber-attack, ranging from phishing to ransomware or serious privacy issues. In order to fight against those ever-evolving threats, Machine Learning is increasingly being used behind the scenes to design better systems that are able of self-learning to boost detection rates and boost overall resilience to unknown attacks. As AI-based solutions penetrate products across the industry, a new kind of threat that is often overlooked is becoming more and more prominent and dangerous: adversarial machine learning (AML).

AML focuses on designing specific inputs to deceive a previously trained Machine Learning models into misclassifying them for a specific purpose. One of the main flaws of any state-of-the-art Machine Learning or Deep Learning algorithms is that they assume that the nature of the data they receive is systematically benign, which is generally the case but does not hold true when an adversarial input is received. The motivation behind altering a ML model into thinking that, for example, a new sample is benign when in fact is malicious can range from pure research to more serious real-life issues such as an autonomous car wrongly classifying a stop sign (and thus provoking a fatal accident) or a wrongly diagnosed disease because of a slightly manipulated magnetic resonance image.

This problem is no exception for Cybersecurity where companies wrongly assume that once the last AI-based product is deployed in their network, their employees are safe...

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.

La integración de RINA con dispositivos basados en el estándar IEEE 802.15.4 permitirá una gestión más eficiente y segura de las redes de sensores, optimizando el rendimiento y la escalabilidad en aplicaciones de Internet de las Cosas. Este proyecto abordará la adaptación de tramas, gestión de recursos, seguridad y calidad de servicio, proporcionando una solución innovadora que mejorará la interoperabilidad y la funcionalidad de los sistemas IoT. La investigación incluirá el diseño, implementación, pruebas y optimización de la shim, con el objetivo de crear una plataforma robusta que potencie las capacidades de RINA en la gestión de redes de sensores y dispositivos inteligentes. La shim 802.15.4 se integrará con la arquitectura RINAsense, el cual es una librería para el rápido prototipado de sensores RINA.

PETGEM is an open-source 3D electromagnetic modeler with support for HPC architectures. It solves Maxwell's equations using a high-order vector finite element method and is primarily written in Python. On the other hand, PETSc is a suite of data structures and routines that provide the building blocks for implementing large-scale application codes on parallel (and serial) computers. PETSc uses the MPI standard for all distributed memory communication.

The main objective of this work is to investigate and implement efficient strategies for the parallel construction of sparse matrices derived from FE analysis within the PETGEM workflow. This development aims to provide fundamental insights into the interoperability and portability of PETGEM by enabling the framework to utilize exascale HPC architectures. The main milestones of the project are:

1. Understand the PETGEM and PETSc basis

2. Perform matrix assembly (in parallel environments)

3. Analyze performance (cache misses, run-time) using Extrae and Paraver (developed at BSC)

4. Implement optimizations (if any)

Given that both projects (PETSc and PETGEM) are under ongoing development, this task necessitates collaboration with both development teams to communicate software engineering requirements, report bugs, and provide patches for issues encountered during development. This collaborative effort will be undertaken with Dr. Stefano Zampini (PETSc developer at KAUST University) and Dr. Octavio Castillo-Reyes (PETGEM developer at UPC/BSC).

For more information, visit: https://www.bsc.es/research-and-development/software-and-apps/software-list/petgem

The main goal of the project is to develop an educational Android app that provides an engaging and educational experience for children, especially those with communication difficulties. The app will integrate with cartoon or child-appropriate video playback, periodically pausing the video to present educational tasks. This feature is designed to encourage active engagement and promote learning during video consumption.

The app will utilize Picture Communication Symbols (PCS) to present questions or challenges visually. Users have to respond to these prompts by selecting the correct PCS image corresponding to the answer. If the answer is correct, the app will resume playback. If the answer is incorrect, the user will have another chance to answer correctly before video playback resumes.

The proposed app's features include:

- Integration a video playback.

- A timer that periodically pauses video playback and presents a task to be solved.

- A database of matching images (such as drawings, photos, and PCS).

- A user-friendly interface that allows adding new matching image pairs to the database, thus customizing the app for individual needs.

- A child-friendly and visually appealing user interface.

The app not only enhances the learning experience for children but also promotes communication skills, cognitive development, and engagement. It provides a valuable tool for parents and caregivers to ensure that screen time is both educational and interactive. Additionally, the app can be customized to suit the individual needs and interests of the child using it. It's important to ensure the app's usability and security. 

In order to improve the app's effectiveness and appropriateness for its target audience, the development of the app will be done in collaboration with professionals in child development and communication disorders. In particular, the project will be partially supervised by members of the TimeAut team (located in Bolzano, Italy; https://www.timeaut.it/) specializing in the treatment of autism and other developmental disorders.

The App is intended for work on a generic Android tablet.