Comunidad Cenia How To
Christ Devia

Christ Devia

Especialidad: Percepción activa, modelos de redes complejas de los estados cerebrales, percepción y memoria de trabajo.
Email: christ.devia@cenia.cl

PUBLICACIONES

[:es]Publisher: eNeuro Link>

ABSTRACT

Variations in human behavior correspond to the adaptation of the nervous system to different internal and environmental demands. Attention, a cognitive process for weighing environmental demands, changes over time. Pupillary activity, which is affected by fluctuating levels of cognitive processing, appears to identify neural dynamics that relate to different states of attention. In mice, for example, pupil dynamics directly correlate with brain state fluctuations. Although, in humans, alpha-band activity is associated with inhibitory processes in cortical networks during visual processing, and its amplitude is modulated by attention, conclusive evidence linking this narrowband activity to pupil changes in time remains sparse. We hypothesize that, as alpha activity and pupil diameter indicate attentional variations over time, these two measures should be comodulated. In this work, we recorded the electroencephalographic (EEG) and pupillary activity of 16 human subjects who had their eyes fixed on a gray screen for 1 min. Our study revealed that the alpha-band amplitude and the high-frequency component of the pupil diameter covariate spontaneously. Specifically, the maximum alpha-band amplitude was observed to occur ∼300 ms before the peak of the pupil diameter. In contrast, the minimum alpha-band amplitude was noted to occur ∼350 ms before the trough of the pupil diameter. The consistent temporal coincidence of these two measurements strongly suggests that the subject’s state of attention, as indicated by the EEG alpha amplitude, is changing moment to moment and can be monitored by measuring EEG together with the diameter pupil.

[:]
RL3 2022
Ir a la publicación

Publisher: Cognitive Systems Research, Elsevier  Link>

ABSTRACT

This article presents a transdisciplinary analysis of the challenges in fusing neuroscience concepts with artificial intelligence (AI) to create AI systems inspired by biological cognition. We explore the structural and functional disparities between the neocortex’s canonical microcircuits and existing AI models, focusing on architectural differences, learning mechanisms, and energy efficiency. The discussion extends to adapting non-goal-oriented learning and dynamic neuronal connections from biological brains to enhance AI’s flexibility and efficiency. This work underscores the potential of neuroscientific insights to revolutionize AI development, advocating for a paradigm shift towards more adaptable and brain-like AI systems. We conclude that there is major room for bioinspiration by focusing on developing architecture, objective functions, and learning rules using a local instead of a global approach.
RL3 2024
Ir a la publicación

Publisher: Elsevier, Expert Systems with Applications  Link>

ABSTRACT

We present a study of an artificial neural architecture that predict human ocular scanpaths while they are free-viewing different images types. This analysis is made by comparing different metrics that encompass scanpath patterns, these metrics aim to measure spatial and temporal errors; such as the MSE, ScanMatch, cross-correlogram peaks, and MultiMatch. Our methodology begin by choosing one architecture and training different parametric models per subject and image type, this allows to adjust the models to each person and a given set of images. We find out that there is a clear difference in prediction when people free-view images with high visual content (high-frequency contents) and low visual content (no-frequency contents). The input features selected for predicting the scanpath are saliency maps calculated from foveated images together with the past of the ocular scanpath of subjects, modeled by our architecture called FovSOS-FSD (Foveated Saliency and Ocular Scanpath with Feature Selection and Direct Prediction).

The results of this study could be used to improve the design of gaze-controlled interfaces, virtual reality, as well as to better understand how humans visually explore their surroundings and pave a way to make future research.

2023
Ir a la publicación

Large language models (LLMs) have revolutionized natural language processing. Understanding their internal mechanisms is crucial for developing more interpretable and optimized architectures. Mechanistic interpretability has led to the development of various methods for assessing layer relevance, with cosine similarity being a widely used tool in the field. On this work, we demonstrate that cosine similarity is a poor proxy for the actual performance degradation caused by layer removal. Our theoretical analysis shows that a layer can exhibit an arbitrarily low cosine similarity score while still being crucial to the model's performance. On the other hand, empirical evidence from a range of LLMs confirms that the correlation between cosine similarity and actual performance degradation is often weak or moderate, leading to misleading interpretations of a transformer's internal mechanisms. We propose a more robust metric for assessing layer relevance: the actual drop in model accuracy resulting from the removal of a layer. Even though it is a computationally costly metric, this approach offers a more accurate picture of layer importance, allowing for more informed pruning strategies and lightweight models. Our findings have significant implications for the development of interpretable LLMs and highlight the need to move beyond cosine similarity in assessing layer relevance.
Investigación Cenia 2026
Ir a la publicación

Understanding visual information processing in the cortex is crucial for neuroscience and visual AI. Studying how different stimuli modify the receptive field characteristics of neurons in the cortex has been a focus for decades. Despite the accepted organization of the visual cortex in columns, understanding how microcircuits process information from various sources remains unclear. Here, we explore the mechanisms and dynamics between microcircuits underlying classical and extra-classical receptive fields effect (ECRF). To do this, we developed a model representing a small portion of the visual cortex consisting of 5 interconnected microcircuit models (4 in V1 and 1 in V2) based on the Potjans model, which describes interactions among excitatory and inhibitory neuron groups across 4 layers of a cortical column (layer 2/3, 4, 5 and 6), and all model parameters were chosen for bio plausibility. We study how long-range lateral connections and top-down interactions from nonstriated cortices modulate V1 neuron activity. To simulate the ECRF effect, we present a preferred stimulus to a microcircuit, followed by the appearance of a new stimulus outside its receptive field. Each stimulus and its characteristics are simulated as an increase in the firing rate of a group of simulated thalamic neurons connected directly to V1. All simulations in our model maintain neuronal activity within reported ranges, indicating stable network parameters. Results show expected neuronal activity during the first stimulus presentation in both complete and modified (without V2) networks. As expected, simulating a stimulus outside the receptive field shows minimal changes in the first microcircuit. Unlike classical effects, simulating the preferred stimulus first and then the outside stimulus shows no significant change in the V2-absent model but suppresses activity in the complete model. Our findings suggest that lateral and vertical connections jointly contribute to the generation of receptive field effects. Furthermore, we noted that achieving the observed dynamics required very fine tuning of the top-down and bottom-up connections between modules, the sensitivity of which could be the origin of all observed effects in the cortex.This work was supported in part by The National Center for Artificial Intelligence CENIA, Chile FB210017.
RL3 2025
Ir a la publicación
agencia nacional de investigación y desarrollo
info@cenia.cl
Edificio de Innovación UC, Piso 2
Vicuña Mackenna 4860
Macul, Chile