arXiv subjects

Jinling Gao

Publications and source records attributed to Jinling Gao.

3 recordsLinked to original sources

arXiv2025

Holistic structure of neural pathways underlies brain perceptual rivalry: Physical mechanism of auditory stream segregation

Brain perceptual rivalry, exemplified by auditory stream segregation of competing tones (A_, B__, ABA_), serves as a core mechanism of brain perception formation. While increasingly recognized as determining by neural connections rather than specific neural groups, the mechanism of brain perception remains uncertain. We demonstrate that auditory stream segregation arises from the topological structure of holistic neural pathways. By constructing a holistic pathway model using existing neurophysiological data, combining nonlinear neural dynamics and nonequilibrium physics, we uncover the biophysical mechanism of perceptual phase transitions from integrated (ABA_) to segregated streams (A_ or B_), as well as the mechanism of temporal dynamics, perceptual switching path, and attention regulation underlying these transitions. Further, we demonstrate how our framework reveals energy consumption of the auditory system and combines it with neuroelectrophysiology. Two psycho-acoustic experiments validate our predictions of perception alternation and attention modulation. Our framework provides a transformative perspective on how brain networks generate complex perceptual experiences, emphasizing the significance of neural pathway structure in the process of brain function realization.

q-bio.NC↗

arXiv2024

PNAS-MOT: Multi-Modal Object Tracking with Pareto Neural Architecture Search

Multiple object tracking is a critical task in autonomous driving. Existing works primarily focus on the heuristic design of neural networks to obtain high accuracy. As tracking accuracy improves, however, neural networks become increasingly complex, posing challenges for their practical application in real driving scenarios due to the high level of latency. In this paper, we explore the use of the neural architecture search (NAS) methods to search for efficient architectures for tracking, aiming for low real-time latency while maintaining relatively high accuracy. Another challenge for object tracking is the unreliability of a single sensor, therefore, we propose a multi-modal framework to improve the robustness. Experiments demonstrate that our algorithm can run on edge devices within lower latency constraints, thus greatly reducing the computational requirements for multi-modal object tracking while keeping lower latency.

cs.CV↗

arXiv2024

G-NAS: Generalizable Neural Architecture Search for Single Domain Generalization Object Detection

In this paper, we focus on a realistic yet challenging task, Single Domain Generalization Object Detection (S-DGOD), where only one source domain's data can be used for training object detectors, but have to generalize multiple distinct target domains. In S-DGOD, both high-capacity fitting and generalization abilities are needed due to the task's complexity. Differentiable Neural Architecture Search (NAS) is known for its high capacity for complex data fitting and we propose to leverage Differentiable NAS to solve S-DGOD. However, it may confront severe over-fitting issues due to the feature imbalance phenomenon, where parameters optimized by gradient descent are biased to learn from the easy-to-learn features, which are usually non-causal and spuriously correlated to ground truth labels, such as the features of background in object detection data. Consequently, this leads to serious performance degradation, especially in generalizing to unseen target domains with huge domain gaps between the source domain and target domains. To address this issue, we propose the Generalizable loss (G-loss), which is an OoD-aware objective, preventing NAS from over-fitting by using gradient descent to optimize parameters not only on a subset of easy-to-learn features but also the remaining predictive features for generalization, and the overall framework is named G-NAS. Experimental results on the S-DGOD urban-scene datasets demonstrate that the proposed G-NAS achieves SOTA performance compared to baseline methods. Codes are available at https://github.com/wufan-cse/G-NAS.

cs.CV↗