Seminarios de Innovación e Investigación en Informática

List of attended talks: - Complex Networks Generation: From Probabilistic Models to Deep Generative Approaches (Jesús Giráldez Crú) - TES X-ray pulse identification using CNNs (Jesús Vega Ferrero) - Realtime analysis of fluorescence imaging data (Andrea Giovanucci) - How to get the best of your scientific communication skills (Carme Roig) - Probabilistic Programming for Inverse Problems in Physical Sciences (Atilim Güne¿ Baydin) - Fuzzy Horn clauses in artificial intelligence: A study of free models, and applications in art painting style categorization (Vicent Costa) - UKRI Centre for Doctoral Training in Safe and Trusted Artificial Intelligence (Natalia Criado Pacheco) - Reinforcement learning in continuous spaces and its applications to robotics (Adrià Colomé Figueras) - Probabilistic Models for Competence Assessment in Education (Alejandra López de Aberasturi Gómez) - The programming architecture of Babbage's Analytical Engine (Raúl Rojas González) - Security and Privacy in AI-based Systems (Jose M Such) - The 6 questions beyond Machine Learning to build successful Data Products (with some lessons learnt from the Yamanas) (Daniel Villatoro) - Brain-Machine Interfaces: Beyond Decoding (José del R. Millán) - Norm Change Under Collective Risk (Giulia Andrighetto) - Recent Advances in Optimal Control and Probabilistic Inference (Vicenç Gómez) - Pushing the Limits of Transformer Models (Santiago Ontañón)

1h - 1h30 / talk

Build and train deep neural networks, identify key architecture parameters, implement vectorized neural networks and deep learning to applications. Train test sets, analyze variance for DL applications, use standard techniques and optimization algorithms, and build neural networks in TensorFlow. Build a CNN and apply it to detection and recognition tasks, use neural style transfer to generate art, and apply algorithms to image and video data. Build and train RNNs, work with NLP and Word Embeddings, and use HuggingFace tokenizers and transformer models to perform NER and Question Answering.

Open

This course is an introduction to the statistical properties of languages and their origins, that are often related to the need of reducing the cognitive effort of speakers or listeners. The course is primarily relevant to anybody interested in how languages (and animal communication) are and why. In this course, students will learn about a myriad of statistical laws of language beyond the scope of traditional courses on information retrieval, how to analyze them and how to explain them based on general principles of human cognition. Students will learn about the mathematical and computational models that have been developed to explain these regularities. During this journey, students will enrich their current knowledge with concepts and tools from linguistics, biology, cognitive science, information theory and multidisciplinary physics. The course is also relevant for researchers interested in evaluating or adapting algorithms, machine learning methods,...based on the real statistical properties of language. As these regularities are often the result of reducing the cognitive effort of language users, the course is also relevant to researchers interested in developing resources or systems that are easier to use or understand by humans natural and language processing tools that exploit the real constraints of the human brain.

to be determined

HackUPC is back! HackUPC is the student hackathon organized by Hackers@UPC since 2016 and, currently, is the largest in Europe! A hackathon is a free event where participants have limited time (in our case 36 hours) to develop a team project from scratch. Apart from the project, there are lots of activities and talks. You'll meet people with your same interests and learn about the newest technologies. Besides, there is the possibility of networking with the sponsoring companies. This year's edition will be hybrid (you can attend either in person or online) and will take place on the weekend of April 29 to May 1 at UPC Campus Nord. We'll provide you with free food, drink, and a place to sleep all weekend long! But that's not all, you'll receive merch and much more! If you don't feel prepared to come as a hacker but still want to participate, apply as a volunteer! You'll feel the hackathon atmosphere and be just like another organizer. To know more about HackUPC and hackathons, you can visit our exposition at the BRGF library from next week until April and attend to our talks there: Wednesday 16th March at 12:30: "What is a hackathon and what is HackUPC." Wednesday 6th April at 12:30: "Pro tips for a hackathon: Introduction to Git, Devpost¿" If you can't come to the talks in person, watch them live on our twitch channel! You can check our social networks and our website to sign up and get more information.

Full days

held as an online event at Jülich Supercomputing Centre (JSC) at Forschungszentrum Jülich. The course was given by Dr. Jan Meinke, Jochen Kreutz, Dr. Kaveh Haghighi-Mood, and Dr. Andreas Herten from JSC as well as by Jiri Kraus and Markus Hrywniak from NVIDIA. GPU-accelerated computing drives current scientific research. Writing fast numeric algorithms for GPUs offers high application performance by offloading compute-intensive portions of the code to an NVIDIA GPU. The course covered basic aspects of GPU architectures and programming. Focus was on the usage of the parallel programming language CUDA-C which allows maximum control of NVIDIA GPU hardware. Examples of increasing complexity were used to demonstrate optimization and tuning of scientific applications.

Desde l'empresa ens han pagat un curs de Bootlin https://bootlin.com (contribuidors al kernel de Linux) sobre Embedded Linux de 24h (https://bootlin.com/doc/training/sessions/hp-online.embedded-linux.mar2022)

9:00-13:00

This tutorial will introduce the audience to the BSC tools for heterogenous programming on FPGA devices. It describes OmpSs@FPGA, as a productive programming environment for compute systems with FPGAs.

EURANOVA, is organising an online 3 days hackathon, called Novhack, from the 25th to the 27th of March. Our goal is to make students live a unique experience through knowledge sharing and coaching, and with real-world data. Concretely, we created Novhack to be: A learning experience: each group will have the opportunity to be accompanied by a Euranovian coach who will guide you and your team throughout this adventure. You will have time to talk to experts who will be able to answer your most technical questions. A concrete project: with our industry partner AISIN, we built a real-life use case, to give you the chance to learn, hands on, the importance of real-time data processing and its use for both business intelligence and advanced analytics. A challenge: you will have to be creative with your group to develop a solution to solve a big data use case that will involve the use of the most popular data streaming technologies. A fun adventure! Filled with teamwork, goodies, great prizes, and unforgettable moments.

All day

The time has come for a New Datathon. This edition we will focus on hate speech detection. As part of the preparation work we are getting inputs and recommendations from members of CNAAE (Comunidad Negra Africana y Afrodescendiente en España) and Sietar España.

Check the website

The tutorial will motivate the audience on the need for portable, efficient programming models that put less pressure on program developers while still getting good performance for clusters and clusters with GPUs.

In the fourth course of the Deep Learning Specialization, you will understand how computer vision has evolved and become familiar with its exciting applications such as autonomous driving, face recognition, reading radiology images, and more. By the end, you will be able to build a convolutional neural network, including recent variations such as residual networks; apply convolutional networks to visual detection and recognition tasks; and use neural style transfer to generate art and apply these algorithms to a variety of image, video, and other 2D or 3D data. The Deep Learning Specialization is our foundational program that will help you understand the capabilities, challenges, and consequences of deep learning and prepare you to participate in the development of leading-edge AI technology. It provides a pathway for you to gain the knowledge and skills to apply machine learning to your work, level up your technical career, and take the definitive step in the world of AI.

Visit to the 4 Years From Now (4YFN) and the Mobile World Congress (MWC) of 2022.

In the workshop with a focus on Deep learning with Keras you will get acquainted with the Keras library and its use through deep learning practical examples. In recent years, deep learning has spread from academic circles and found its place in industry, advertising, healthcare, banking, security etc. In the workshop the most useful neural network architectures and their components will be presented. The different models will be implemented within the Colaboratory cloud service, which includes all the necessary software and enables easy code sharing and execution on Google's high-performance infrastructure (GPU, TPU). Through examples, you will be acquainted with the main types of data structures for machine learning and how we use them in conjunction with neural networks. We will present guidelines for designing an appropriate model and the common obstacles that encounter in deep learning algorithms

Embedded Software and Hardware Architecture is a first dive into understanding embedded architectures and writing software to manipulate this hardware. You will gain experience writing low-level firmware to directly interface hardware with highly efficient, readable and portable design practices. We will now transition from the Host Linux Machine where we built and ran code in a simulated environment to an Integrated Development Environment where you will build and install code directly on your ARM Cortex-M4 Microcontroller. Course assignments include writing firmware to interact and configure both the underlying ARM architecture and the MSP432 microcontroller platform. The course concludes with a project where you will develop a circular buffer data structure.

This third online workshop gathers leading researchers in high-performance computing from the JLESC partners INRIA, the University of Illinois, Argonne National Laboratory, Barcelona Supercomputing Center, Jülich Supercomputing Centre, RIKEN Center for Computational Science and the University of Tennessee to explore the most recent and critical issues in advancing the field of HPC from petascale to the extreme scale era. The workshop will feature short talk and break-out sessions on these seven central topics: - Artificial intelligence - Big Data, I/O and in-situ visualization - Numerical methods and algorithms - Resilience - Performance tools - Programming Languages - Advanced architectures - A key objective of the workshop is to foster the existing collaborations and inform the workshop attendees of their respective progress, as well as identify new research collaborations and establish a roadmap for their implementation. The workshop is open to all participatants from the JLESC institutions Illinois, INRIA, ANL, BSC, JSC, Riken R-CCS and UTK; faculties, researchers, engineers and students who want to learn more about Post-Petascale / Pre-Exascale Computing. Contact your local executive director if you are interested in attending.

08:00-11:15 ET

Hackathon for the La Marato TV3 i Catalunya Informació, for the Mental Health

The RISC-V Summit is the annual conference organised by the RISC-V Foundation, featuring 3 days of keynotes, technical talks and tutorials in addition to networking opportunities. In this year's event which will take place online, there are 1 speaker from UPC/BSC: Roger Espasa.

This hackathon targets application developers that are interested in investigating the impact of novel SVE extensions in their HPC applications.

I have attended the full week of MICRO 2021, i.e. 3 full days of talks and paper presentations and 2 days of tutorials/workshops.

https://www.microarch.org/micro54/program/

ICCV is the premier international computer vision event comprising the main conference and several co-located workshops and tutorials. With its high quality and low cost, it provides an exceptional value for students, academics and industry researchers.

https://iccv2021.thecvf.com/overview

This is a multi-month hardware security contest. The organisers provide the RTL code of a "buggy" SoC (last year it was the RISC-V Ariane with some vulnerabilities injected) together with a security specification. The participants, who can be from industry or students, are organised in teams of 4 under the supervision of an academic advisor and they have to perform security verification, so they need to find, fix and report as many security vulnerabilities as they can. The top teams of the first phase will be invited to compete in a similar, one/two days event at DAC (if the conference will be organised in a physical form) and receive monetary prizes. In that case, the travel cost will be covered by a travel grant from the competition sponsors and the local organiser from UPC will accompany the qualified teams.

Self paced project development and weekly meetings with the project advisor.

This is a multi-month hardware security contest. The organisers provide the RTL code of a "buggy" SoC (last year it was the RISC-V Ariane with some vulnerabilities injected) together with a security specification. The participants, who can be from industry or students, are organised in teams of 4 under the supervision of an academic advisor and they have to perform security verification, so they need to find, fix and report as many security vulnerabilities as they can. The top teams of the first phase will be invited to compete in a similar, one/two days event at DAC (if the conference will be organised in a physical form) and receive monetary prizes. In that case, the travel cost will be covered by a travel grant from the competition sponsors and the local organiser from UPC will accompany the qualified teams.

Self paced project development and weekly meetings with the project advisor.

The project consists in finding a rectangle-based approximation of a boolean matrix using a pre-defined number of rectangles. This is a combinatorial optimization problem that helps to discretize the floorplanning problem from a continuous representation of the components to a rectangle-based representation.

The ACACES Summer School is organized by the HiPEAC Network in collaboration with the TETRAMAX Innovation Action. ACACES 2021 is the seventeenth edition of the Summer School. The ACACES Summer School is a one week summer school for computer architects and tool builders working in the field of high performance computer architecture and compilation for computing systems. The school aims at the dissemination of advanced scientific knowledge and the promotion of international contacts among scientists from academia and industry. A distinguishing feature of this Summer School is its broad scope ranging from low level technological issues to advanced compilation techniques. In the design of modern computer systems one has to be knowledgeable about architecture as well as about the quality of the code, and how to improve it. This summer school offers the ideal mix of the two worlds ¿ both at the entry level and at the most advanced level. The summer school is open to everybody but previous training and/or experience in computer science as well as a background in computer architecture or compilation is indispensable.

Daily full day lectures

ACM Europe Summer School consists of a series of talks and lectures as well as practical exercises hosted by the Barcelona Supercomputing Center (BSC-CNS) and the Universitat Politècnica de Catalunya (UPC). Prof. Luca Cardelli (University of Oxford) will give a keynote lecture on Molecular Computing. Uri Weiser (Technion) will address state-of-the-art architectures for AI, followed by two sessions dedicated to the design of RISC V accelerators, by Luca Benini (ETHZ and University of Bologna) and Jesús Labarta (BSC). Jordi Torres and Josep Lluis Berral (BSC and UPC) will cover topics related to software stack for AI, by addressing deep learning and distributed data analytics in supercomputing systems. The closing session of the school will include a lecture by Mauro Olivieri (Sapienza University) and it will be dedicated to applications for AI, focused on Edge computing and vector acceleration support.

9:00-18:00

Learn how to create and mod video games using Unreal Engine 4, even if you're a complete beginner. Unreal is a free-to-use game development engine used by AAA studios and indie developers worldwide. It can be a complex beast to get into, but we break it down step-by-step

22/03/2021. Big Data: background and defintions. Least squares: background and the Gauss-Markov Theorem. Lp-norm estimators and the Exponential Power Function that is the Generalized Error Distribution. 23/03/2021. Main characteristics of Big Data and its basic management. Important suggestions are developed for a quality control before to extract significant samples for subsequent analysis. 24/03/2021. Big Data applications: Big Data and forensics: an innovative approach for a predictable jurisprudence. Big Data for Social and Economics Science. 25/03/2021. Introduction to Exponential Power Function (E.P.F.) and Lp¿norm estimators. Comparison of the Lp¿norm estimators with the Least Squares method (L2) by simulation studies, introducing a linear regression model in the case of multicollinearity. A comparative analysis applying three estimation methods to evaluate the empirical distribution of the regression parameters is considered. 26/03/2021. Practice of the usage of the normalp package of R. Test of fit-of- godness of the model we present. R environment has a set of tools for dealing with the Exponential Power Function. In this package there are functions to compute the density function, the distribution function and the quantiles from an Exponential Power Function and to generate pseudo-random numbers from the same distribution. 29/03/2021. A Generalized Error Distribution Copula-based method for portfolios risk assessment. Comparison of G.E.D. with Gaussian distribution to financial markets and examples. Applications with R. 30/03/2021. Clustering Time Series using the Skewed Generalized Error Distribution (S.G.E.D): a fuzzy approach based on higher conditional moments. 31/03/2021. Discussion and comments on the collective project titled:"Add a new distribution (S.G.E.D) to the GAS package using R studio". Interactive lesson with final evaluation.

22/03/2021. Big Data: background and defintions. Least squares: background and the Gauss-Markov Theorem. Lp-norm estimators and the Exponential Power Function that is the Generalized Error Distribution. 23/03/2021. Main characteristics of Big Data and its basic management. Important suggestions are developed for a quality control before to extract significant samples for subsequent analysis. 24/03/2021. Big Data applications: Big Data and forensics: an innovative approach for a predictable jurisprudence. Big Data for Social and Economics Science. 25/03/2021. Introduction to Exponential Power Function (E.P.F.) and Lp¿norm estimators. Comparison of the Lp¿norm estimators with the Least Squares method (L2) by simulation studies, introducing a linear regression model in the case of multicollinearity. A comparative analysis applying three estimation methods to evaluate the empirical distribution of the regression parameters is considered. 26/03/2021. Practice of the usage of the normalp package of R. Test of fit-of- godness of the model we present. R environment has a set of tools for dealing with the Exponential Power Function. In this package there are functions to compute the density function, the distribution function and the quantiles from an Exponential Power Function and to generate pseudo-random numbers from the same distribution. 29/03/2021. A Generalized Error Distribution Copula-based method for portfolios risk assessment. Comparison of G.E.D. with Gaussian distribution to financial markets and examples. Applications with R. 30/03/2021. Clustering Time Series using the Skewed Generalized Error Distribution (S.G.E.D): a fuzzy approach based on higher conditional moments. 31/03/2021. Discussion and comments on the collective project titled:"Add a new distribution (S.G.E.D) to the GAS package using R studio". Interactive lesson with final evaluation.

In this course, you will learn how to set up and use a Vulkan application, as well as implement several key features exclusive to Vulkan itself, such as Descriptor Sets, Push Constants, and Subpasses. Each aspect is explained in great amounts of detail, to ensure you really understand what all the parts mean, and how they all connect together.

Build and train deep neural networks, identify key architecture parameters, implement vectorized neural networks and deep learning to applications. Train test sets, analyze variance for DL applications, use standard techniques and optimization algorithms, and build neural networks in TensorFlow. Build a CNN and apply it to detection and recognition tasks, use neural style transfer to generate art, and apply algorithms to image and video data. Build and train RNNs, work with NLP and Word Embeddings, and use HuggingFace tokenizers and transformer models to perform NER and Question Answering.