Speaker: Miguel Ángel Martínez Pay. Abstract: This work presents a comprehensive overview of Large Language Models (LLMs), from their theoretical framework to practical applications in text classification. It compares the effectiveness of two key approaches: fine-tuning embeddings of smaller models… Read More
Integration of Emotional Information in Speaker Recognition Systems
Speaker: Arturo Domínguez Santos. Abstract: This Master’s thesis addresses the challenge of investigating how emotions affect speakerverification and proposes a system that integrates this emotional variability to try toimprove accuracy. The focus is on the speaker’s emotions, which has traditionally… Read More
Exploring Speech Foundation Models for End-to-End Speaker Diarization
Speaker: Laura Herrera Alarcón. Abstract: In this Master’s Thesis the use of pre-trained models for the diarization task has beenstudied in order to exploit their ability to extract robust and discriminative features.In particular, the WavLM model has been combined with… Read More
Interpretation of fingerprint evidence with likelihood ratios (LRs – Likelihood ratios)
Speaker: Joaquín González Rodríguez. Abstract: The forensic fingerprint identification process based on the ACE-V method, widely implemented, makes absolute identification or exclusion decisions that depend on opinions that vary from expert to expert (for example, whether we consider an observed… Read More
SGE & CCC Architecture – Introduction for Beginners
Speaker: Adrián Aranda Márquez. Abstract: Simple technical introduction for using SGE with AUDIAS servers (Son of Grid Engine) and CCC (UAM’s Central Computing Center).
Stabilising Reinforcement Learning with Past Action-State Representation Learning
Speaker: Tamas Endrei. Abstract: Although deep reinforcement learning (DRL) deals with sequential decision-making problems, temporal information representation is absent from state-of-the-art actor-critic algorithms. The reliance on only the current timestep information causes instability in concurrent actions. Furthermore, the over-reliance on… Read More
Study of the predictive capacity of the efficacy of platelet-rich plasma (PRP) treatments in joint injuries
Speaker: Berta Caunedo Castro. Abstract: This Final Degree Project evaluates Platelet-Rich Plasma (PRP) therapy as an alternative to traditional treatments for knee osteoarthritis, a prevalent joint condition. PRP uses regenerative growth factors from the patient’s blood, but its variability complicates… Read More
Enhancing Sound Event Detection and Speaker Verification employing weak supervision
Speaker: Sara Barahona Quirós. Abstract: In this seminar, we will explore approaches for training acoustic event detection and speaker verification systems employing limited labels. Specifically, for the first task, we will explain the optimization process of a system based on… Read More
Leveraging Speaker Embeddings in End-to-End Neural Diarization for Two-Speaker Scenarios
Speaker: Juan Ignacio Álvarez Trejos. Abstract: This presentation covers the work presented at Odyssey 2024, focusing on speaker diarization in two-speaker scenarios. End-to-end neural speaker diarization systems are designed to handle overlapping speech while accurately distinguishing between speakers. In this… Read More
Transformers for Binding Prediction of Hypoxia-Induced Factors
Speaker: Manuel Fernando Mollón Laorca. Abstract: Hypoxia-inducible factors (HIFs) are proteins that play a crucial role in the cellular response to low oxygen levels. Accurate prediction of the binding of these factors to their target DNA is essential for understanding… Read More