Research overview

Cognitive and Computational Neuroscience

Speech and Language Processing

• Bridging Brains and Machines: Investigating how speech and language are represented and processed in the human brain and AI systems, with a focus on Large Language Models (LLMs).

• Continuous Speech Perception: Using advanced neuroimaging techniques such as MEG, EEG, and invasive iEEG to study natural speech (e.g., audiobooks) and uncover complex brain functions.

• Auditory Phantom Perceptions: Studying phenomena like tinnitus and hyperacusis to advance understanding of pathological speech and hearing disorders.

Cognitive Maps

• Internal Representations: Modeling how the brain organizes and navigates thoughts and complex information, such as linguistic structures, through neural network-based successor representations.

• Applications: Exploring scenarios like spatial navigation, semantic abstraction, and hierarchical language processing by integrating multi-scale representations, Bayesian frameworks, and predictive coding principles.

Brain-Constrained AI Models

• Deep Neural Networks: Developing biologically plausible models to simulate brain functions, advancing robust, explainable AI inspired by neuroscience.

Theoretical Neuroscience and Cognitive Physics

Recurrent Neural Networks (RNNs)

• Investigating the structural and dynamic properties of RNNs using theoretical physics methods like dynamical systems, chaos theory, and information theory.

• Exploring the role of noise and randomness in neural information processing.

Bayesian Brain and Predictive Coding

• Understanding how auditory and speech perceptions emerge from interactions between top-down predictive coding and bottom-up adaptive stochastic resonance.

• Advancing the Bayesian brain framework to study healthy and pathological speech processing.

Machine Learning and Data Science

Natural Language Processing

• Continuous Speech and Neural Dynamics: Using NLP tools to align continuous speech stimuli with neural recordings, enabling real-time analysis of linguistic processing in the brain.

• Linguistic Complexity and Neural Patterns: Investigating neural responses to varying linguistic structures (e.g., words, phonemes) using NLP-based segmentation and forced alignment techniques.

• Corpus-Driven Neurolinguistics: Integrating computational corpus linguistics with neuroimaging to explore real-world language processing and generate hypotheses from large-scale datasets.

Neuroimaging Data Analysis

• Deep Learning and Clustering: Developing cutting-edge methodologies to analyze and visualize high-dimensional, multi-modal neuroimaging data.

• NLP driven Neuroimaging Analysis: Applying NLP frameworks to enhance the precision of neural data synchronization and the study of linguistic phenomena in naturalistic conditions. 

Explainable AI and NeuroAI

• Demystifying Neural Networks: Addressing AI’s black box problem by reverse-engineering and interpreting the behavior of neural networks, including LLMs.

• Principle Transfer: Translating neuroscience principles into AI to design efficient, robust, and interpretable architectures.

My vision

Through this work, I aim to uncover the fundamental mechanisms underlying human cognition and perception while translating these insights into AI systems that are not only powerful but also comprehensible and human-aligned.