“I supervised Alberto's M.Sc. thesis for the Artificial Intelligence program at my department. He addressed a difficult problem (recommendation in social networks from very incomplete information), and obtained results and insights beyond what imagined. More importantly, during that time Alberto exhibited unusual initiative, resources, goal-oriented mentality, and highest professional ethics. I recommend him heartily.”
About
Activity
-
How important is Online Data for RLHF? 👀 HyPO or Hybrid Preference Optimization is another example of how important online/on-policy Data is to…
How important is Online Data for RLHF? 👀 HyPO or Hybrid Preference Optimization is another example of how important online/on-policy Data is to…
Liked by Alberto Lumbreras
-
🤖 Bones notícies per l'ecosistema català d'IA. Catalunya s’integra a la xarxa d’excel·lència en recerca i innovació sobre #IA, European Laboratory…
🤖 Bones notícies per l'ecosistema català d'IA. Catalunya s’integra a la xarxa d’excel·lència en recerca i innovació sobre #IA, European Laboratory…
Liked by Alberto Lumbreras
-
A very good and complete overview from Chip Huyen on how people are building Generative AI platforms these days. It's therapeutic to see that…
A very good and complete overview from Chip Huyen on how people are building Generative AI platforms these days. It's therapeutic to see that…
Shared by Alberto Lumbreras
Experience & Education
Licenses & Certifications
Volunteer Experience
-
Volunteer
Associació d'Amistat amb el Poble de Guatemala
- Present 18 years 10 months
Human Rights
• Guatemala.
• Collaboration with local and international organizations (Red Cross) in the context of the Hurricane Stan crisis. Distribution of emergency aid in rural areas.
• http://www.aapguatemala.org/05_brigades/index_05_brigades.htm -
Volunteer
Fundació PROIDE
- Present 18 years 1 month
Human Rights
• Togo.
• Participation in the construction of a house for a cooperative of local farmers.
• Installation of solar pumps for wells.
• http://tinyurl.com/proide-togo -
Volunteer
Fundació PROIDE
- Present 17 years
Human Rights
• Madagascar.
• Participation in the painting of a school in the city of Ambrosita.
• Installation of LAN network for the school computer lab.
• http://tinyurl.com/proide-madagascar -
Volunteer
Associació d'Amistat amb el Poble de Guatemala
- Present 16 years
Human Rights
• Guatemala.
• Participation in alphabetization campaign.
• Support to local organizations such as farmer unions.
Publications
-
LOCOST: State-Space Models for Long Document Abstractive Summarization.
Association for Computational Linguistics
State-space models are a low-complexity alternative to transformers for encoding long sequences and capturing long-term dependencies. We propose LOCOST: an encoder-decoder architecture based on state-space models for conditional text generation with long context inputs. With a computational complexity of 𝒪(L log L), this architecture can handle significantly longer sequences than state-of-the-art models that are based on sparse attention patterns. We evaluate our model on a series of long…
State-space models are a low-complexity alternative to transformers for encoding long sequences and capturing long-term dependencies. We propose LOCOST: an encoder-decoder architecture based on state-space models for conditional text generation with long context inputs. With a computational complexity of 𝒪(L log L), this architecture can handle significantly longer sequences than state-of-the-art models that are based on sparse attention patterns. We evaluate our model on a series of long document abstractive summarization tasks. The model reaches a performance level that is 93-96% comparable to the top-performing sparse transformers of the same size while saving up to 50% memory during training and up to 87% during inference. Additionally, LOCOST effectively handles input texts exceeding 600K tokens at inference time, setting new state-of-the-art results on full-book summarization and opening new perspectives for long input processing.
Other authorsSee publication -
Learning from Multiple Sources for Data-to-Text and Text-to-Data
International Conference on Artificial Intelligence and Statistics
Data-to-text (D2T) and text-to-data (T2D) are dual tasks that convert structured data, such as graphs or tables into fluent text, and vice versa. These tasks are usually handled separately and use corpora extracted from a single source. Current systems leverage pre-trained language models fine-tuned on D2T or T2D tasks. This approach has two main limitations: first, a separate system has to be tuned for each task and source; second, learning is limited by the scarcity of available corpora. This…
Data-to-text (D2T) and text-to-data (T2D) are dual tasks that convert structured data, such as graphs or tables into fluent text, and vice versa. These tasks are usually handled separately and use corpora extracted from a single source. Current systems leverage pre-trained language models fine-tuned on D2T or T2D tasks. This approach has two main limitations: first, a separate system has to be tuned for each task and source; second, learning is limited by the scarcity of available corpora. This paper considers a more general scenario where data are available from multiple heterogeneous sources. Each source, with its specific data format and semantic domain, provides a non-parallel corpus of text and structured data. We introduce a variational autoencoder model with disentangled style and content variables that allows us to represent the diversity that stems from multiple sources of text and data. Our model is designed to handle the tasks of D2T and T2D jointly. We evaluate our model on several datasets, and show that by learning from multiple sources, our model closes the performance gap with its supervised single-source counterpart and outperforms it in some cases.
Other authorsSee publication -
Bayesian Mean-parameterized Nonnegative Binary Matrix Factorization
Data Mining and Knowledge Discovery
Binary data matrices can represent many types of data such as social networks, votes, or gene expression. In some cases, the analysis of binary matrices can be tackled with nonnegative matrix factorization (NMF), where the observed data matrix is approximated by the product of two smaller nonnegative matrices. In this context, probabilistic NMF assumes a generative model where the data is usually Bernoulli-distributed. Often, a link function is used to map the factorization to the [0, 1] range,…
Binary data matrices can represent many types of data such as social networks, votes, or gene expression. In some cases, the analysis of binary matrices can be tackled with nonnegative matrix factorization (NMF), where the observed data matrix is approximated by the product of two smaller nonnegative matrices. In this context, probabilistic NMF assumes a generative model where the data is usually Bernoulli-distributed. Often, a link function is used to map the factorization to the [0, 1] range, ensuring a valid Bernoulli mean parameter. However, link functions have the potential disadvantage to lead to uninterpretable models. Mean-parameterized NMF, on the contrary, overcomes this problem. We propose a unified framework for Bayesian mean-parameterized nonnegative binary matrix factorization models (NBMF). We analyze three models which correspond to three possible constraints that respect the mean-parameterization without the need for link functions. Furthermore, we derive a novel collapsed Gibbs sampler and a collapsed variational algorithm to infer the posterior distribution of the factors. Next, we extend the proposed models to a nonparametric setting where the number of used latent dimensions is automatically driven by the observed data. We analyze the performance of our NBMF methods in multiple datasets for different tasks such as dictionary learning and prediction of missing data. Experiments show that our methods provide similar or superior results than the state of the art, while automatically detecting the number of relevant components.
Other authorsSee publication -
Closed-form Marginal Likelihood in Gamma-Poisson Matrix Factorization
International Conference on Machine Learning
We present novel understandings of the Gamma-Poisson (GaP) model, a probabilistic matrix factorization model for count data. We show that GaP can be rewritten free of the score/activation matrix. This gives us new insights about the estimation of the topic/dictionary matrix by maximum marginal likelihood estimation. In particular, this explains the robustness of this estimator to over-specified values of the factorization rank and in particular its ability to automatically prune spurious…
We present novel understandings of the Gamma-Poisson (GaP) model, a probabilistic matrix factorization model for count data. We show that GaP can be rewritten free of the score/activation matrix. This gives us new insights about the estimation of the topic/dictionary matrix by maximum marginal likelihood estimation. In particular, this explains the robustness of this estimator to over-specified values of the factorization rank and in particular its ability to automatically prune spurious dictionary columns, as empirically observed in previous work. The marginalization of the activation matrix leads in turn to a new Monte-Carlo Expectation-Maximization algorithm with favorable properties.
Other authorsSee publication -
Role detection in online forums based on growth models for trees
Social Network Analysis and Mining
Some structural characteristics of online discussions have been successfully modeled in the recent years. When parameters of these models are properly estimated, the models are able to generate synthetic discussions that are structurally similar to the real discussions. A common aspect of these models is that they consider that all users behave according to the same model. In this paper, we combine a growth model with an Expectation–Maximization algorithm that finds different parameters for…
Some structural characteristics of online discussions have been successfully modeled in the recent years. When parameters of these models are properly estimated, the models are able to generate synthetic discussions that are structurally similar to the real discussions. A common aspect of these models is that they consider that all users behave according to the same model. In this paper, we combine a growth model with an Expectation–Maximization algorithm that finds different parameters for different latent groups of users. We use this method to find the different roles that coexist in the community. Moreover, we analyze whether we can predict users behaviors based on their roles. Indeed, we show that predictions are improved for some of the roles when compared with a simple growth model.
Other authorsSee publication -
Non-parametric clustering over user features and latent behavioral functions with dual-view mixture models
Computational Statistics
We present a dual-view mixture model to cluster users based on their features and
latent behavioral functions. Every component of the mixture model represents a probability
density over a feature view for observed user attributes and a behavior view for latent behavioral
functions that are indirectly observed through user actions or behaviors. Our task is to infer the
groups of users as well as their latent behavioral functions. We also propose a non-parametric
version based on a…We present a dual-view mixture model to cluster users based on their features and
latent behavioral functions. Every component of the mixture model represents a probability
density over a feature view for observed user attributes and a behavior view for latent behavioral
functions that are indirectly observed through user actions or behaviors. Our task is to infer the
groups of users as well as their latent behavioral functions. We also propose a non-parametric
version based on a Dirichlet Process to automatically infer the number of clusters. We test the
properties and performance of the model on a synthetic dataset that represents the participation
of users in the threads of an online forum. Experiments show that dual-view models outperform
single-view ones when one of the views lacks information.Other authorsSee publication -
Analyse des rôles dans les communautés virtuelles : définitions et premières expérimentations sur IMDb
MARAMI 2013
Analyser les rôles dans les communautés virtuelles nous permet de mieux comprendre, voire de prédire, le comportement individuel des internautes. Bien que de nombreuses approches aient été proposées, on constate un manque de généralisation des méthodes existantes et des résultats obtenus. Dans ce papier, nous passons en revue quelques théories développées à propos des rôles sociaux et nous cherchons une définition compatible à une automatisation par les machines de la détection des rôles joués…
Analyser les rôles dans les communautés virtuelles nous permet de mieux comprendre, voire de prédire, le comportement individuel des internautes. Bien que de nombreuses approches aient été proposées, on constate un manque de généralisation des méthodes existantes et des résultats obtenus. Dans ce papier, nous passons en revue quelques théories développées à propos des rôles sociaux et nous cherchons une définition compatible à une automatisation par les machines de la détection des rôles joués par les individus dans des fils de discussions sur internet. Nous analysons ensuite le site Web IMDb afin d’illustrer notre discours.
Other authorsSee publication -
Tecnopolítica: la potencia de las multitudes conectadas. El sistema red 15M, un nuevo paradigma de la política distribuida.
IN3 Working Paper Series
El estudio recogido en este documento corresponde a una experiencia de investigación colectiva del grupo @datAnalysis15M sobre el movimiento #15M, también denominado movimiento de l@s indignad@s o #spanishrevolution. El estudio, apoyado en la experiencia interna del movimiento y en literatura previa para establecer el marco teórico, se ha realizado a través de distintos métodos experimentales de análisis de datos, desde el análisis de redes sociales y de sistemas complejos al procesamiento de…
El estudio recogido en este documento corresponde a una experiencia de investigación colectiva del grupo @datAnalysis15M sobre el movimiento #15M, también denominado movimiento de l@s indignad@s o #spanishrevolution. El estudio, apoyado en la experiencia interna del movimiento y en literatura previa para establecer el marco teórico, se ha realizado a través de distintos métodos experimentales de análisis de datos, desde el análisis de redes sociales y de sistemas complejos al procesamiento de lenguaje y el análisis de emociones. La investigación atiende a los factores de antecedentes, gestación y desencadenantes, al mismo tiempo que muestra la explosión y desarrollo del #15M. También se explica la importancia de la relación entre la multiplicación de las prácticas tecnopolíticas a través de redes sociales humanas y digitales y las prácticas resultantes de organización, acción y comunicación colectiva. En el estudio se consideran la toma del espacio urbano y la explosión emocional de indignación y empoderamiento que, junto a estas prácticas, dan lugar a un sistema-red autónomo y autoorganizado, en otras palabras, una multitud conectada inteligente.
Other authorsSee publication -
Applying trust metrics based on user interactions to recommendation in social networks
1st International Workshop of Social Knowledge Discovery and Utilization
Recommender systems have been strongly researched within the last decade. With the arising and popularization of digital social networks a new field has been opened for social recommendations. Considering the network topology, users interactions, or estimating trust between users are some of the new strategies that recommender systems can take into account in order to adapt their techniques to these new scenarios. We introduce MarkovTrust, a way to infer trust from Twitter interactions and to…
Recommender systems have been strongly researched within the last decade. With the arising and popularization of digital social networks a new field has been opened for social recommendations. Considering the network topology, users interactions, or estimating trust between users are some of the new strategies that recommender systems can take into account in order to adapt their techniques to these new scenarios. We introduce MarkovTrust, a way to infer trust from Twitter interactions and to compute trust between distant users. MarkovTrust is based on Markov chains, which makes it simple to be implemented and computationally efficient. We study the properties of this trust metric and study its application in a recommender system of tweets.
Other authorsSee publication
Projects
-
EU Projects proposals
-
Contributed to several proposals writings for the 6th and 7th European Framework Program, related to social networks and peer-to-peer.
-
Recommendations and user profiling
-
Worked on the development of a user profiling platform back-end.
-
Multi-platform casual gamining (spin-off)
-
Worked in the formation of a start-up founded by Telefonica in collaboration with the Center for Digital Media (Vancouver, http://thecdm.ca/). The goal of the start-up was to bring casual games into any mobile device. Convergence PC-mobile, mobile platform agnosticism, and a play for free business model were the main points of the project.
http://thecdm.ca/projects/archives/movisphere -
Apps crowdsourcing plafform
-
Lead a development team in a project to develop a crowdsourcing programming platform. The platform aimed to achieve a large amount of widgets programmed by users and automatically adapted to PC, mobile, and TV.
-
Campus Party
-
Campus Party is the biggest LAN Party in Spain. On behalf of Telefonica R&D I coordinate the activities which take place in the event, workshops, seminars, and programming contests which provided interesting inputs to Telefónica projects as well as an opportunity to recruit students and junior engineers.
Our participation in the…Campus Party is the biggest LAN Party in Spain. On behalf of Telefonica R&D I coordinate the activities which take place in the event, workshops, seminars, and programming contests which provided interesting inputs to Telefónica projects as well as an opportunity to recruit students and junior engineers.
Our participation in the news:
http://www.20minutos.es/noticia/403790/0/talentos/campus/party/
http://www.ccma.cat/324/Els-cercatalents-busquen-joves-promeses-a-la-Campus-Party/noticia/299356/
http://www.elconfidencial.com/sociedad/2008-07-31/microsoft-telefonica-y-el-ejercito-rastrean-la-campus-en-busca-de-talentos_359552/ -
Video and mobile P2P
-
Mobile P2P: Development of a P2P system for rapid viral content distribution over Bluetooth connections.
Video P2P: Prospective project to analyze the state of the art of P2P-TV technologies, their possible deployment in Telefonica multiplatform network, and new business prospection. -
Expert systems
-
Design and development of an expert system that monitors and automatically corrects failures in a tier center of the EGEE network (Enabling Grids for E-Science in Europe)
Honors & Awards
-
Best Paper Award at the 18th Conference of the European Chapter of the Association for Computational Linguistics
Association for Computational Linguistics
Best Paper Award to the paper "LOCOST: State-Space Models for Long Document Abstractive Summarization" https://aclanthology.org/2024.eacl-long.69/
-
Best New Business Opportunity
Telefónica
“Movisphere” is a multi-platform, virtual world arcade accessible by and between computer and mobile technology. The game won top honours in the "Best New Business Opportunity" Category at Telefonica's Second Annual Research Fair.
http://thecdm.ca/projects/archives/movisphere -
Honoree at Integrated Mobile Project
Webby Awards
My role was to co-lead the team of developers that would create the prototype in Vancouver, in collaboration with the Center of Digital Media, and to elaborate a business plan for a Telefónica spin-off.
http://webbyawards.com/winners/2009/mobile-apps/general-mobile-web-categories/integrated-mobile-experience/movisphere/
Languages
-
Inglés
Full professional proficiency
-
Spanish
Native or bilingual proficiency
-
Catalan
Native or bilingual proficiency
-
French
Professional working proficiency
Recommendations received
2 people have recommended Alberto
Join now to viewMore activity by Alberto
-
Building a platform for generative AI applications Link: https://lnkd.in/gDUcwQ-v After studying how companies deploy generative AI applications, I…
Building a platform for generative AI applications Link: https://lnkd.in/gDUcwQ-v After studying how companies deploy generative AI applications, I…
Liked by Alberto Lumbreras
-
Meta just released Llama 3.1, a new "open-source model". While I appreciate the huge contribution from Meta LLMs, ✅ the weights are available ✅…
Meta just released Llama 3.1, a new "open-source model". While I appreciate the huge contribution from Meta LLMs, ✅ the weights are available ✅…
Shared by Alberto Lumbreras
-
Before too many ad tech companies start celebrating the extension in the availability of third-party cookies in Chrome, it is important to await the…
Before too many ad tech companies start celebrating the extension in the availability of third-party cookies in Chrome, it is important to await the…
Liked by Alberto Lumbreras
-
🚨 Google ne va finalement pas supprimer les cookies tiers de Chrome 🚨 Après des années à repousser l'échéance, le géant de la publicité fait cette…
🚨 Google ne va finalement pas supprimer les cookies tiers de Chrome 🚨 Après des années à repousser l'échéance, le géant de la publicité fait cette…
Liked by Alberto Lumbreras
-
It's another crazy week for open AI so here's a brief recap of what happened! - Hugging Face released SmolLM, a series of super tiny but powerful…
It's another crazy week for open AI so here's a brief recap of what happened! - Hugging Face released SmolLM, a series of super tiny but powerful…
Liked by Alberto Lumbreras
-
Smol but mighty! 💪 Hugging Face's research team just released SmolLM, 3 super fast models that can run virtually everywhere. These SOTA models: -…
Smol but mighty! 💪 Hugging Face's research team just released SmolLM, 3 super fast models that can run virtually everywhere. These SOTA models: -…
Liked by Alberto Lumbreras
-
In just a few years, the majority of AI usage might happen on-Device. Excited to share SmolLM a series of small language models from 135M, 360M, and…
In just a few years, the majority of AI usage might happen on-Device. Excited to share SmolLM a series of small language models from 135M, 360M, and…
Liked by Alberto Lumbreras
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Alberto Lumbreras in Spain
-
Alberto Lumbreras
at Roche
-
Alberto Lumbreras
Estudiante en Universidad Complutense de Madrid
-
Alberto Lumbreras
Tecnico en Construcciones Metálicas
-
Alberto Ibáñez Lumbreras
Programador analítico en Efedos Consulting Empresarial, S.L.
17 others named Alberto Lumbreras in Spain are on LinkedIn
See others named Alberto Lumbreras