Search for: All records

Learning from noisy labels is a long-standing problem in machine learning for real applications. One of the main research lines focuses on learning a label corrector to purify potential noisy labels. However, these methods typically rely on strict assumptions and are limited to certain types of label noise. In this paper, we reformulate the label-noise problem from a generative-model perspective, i.e., labels are generated by gradually refining an initial random guess. This new perspective immediately enables existing powerful diffusion models to seamlessly learn the stochastic generative process. Once the generative uncertainty is modeled, we can perform classification inference using maximum likelihood estimation of labels. To mitigate the impact of noisy labels, we propose Label-Retrieval- Augmented (LRA) diffusion model 1, which leverages neighbor consistency to effectively construct pseudo-clean labels for diffusion training. Our model is flexible and general, allowing easy incorporation of different types of conditional information, e.g., use of pre-trained models, to further boost model performance. Extensive experiments are conducted for evaluation. Our model achieves new state-of-the-art (SOTA) results on all standard real-world benchmark datasets. Remarkably, by incorporating conditional information from the powerful CLIP model, our method can boost the current SOTA accuracy by 10-20 absolute points in many cases. 

A widely used cerebrovascular stimulus and common pathophysiologic condition, hypercapnia is of great interest in brain research. However, it remains controversial how hypercapnia affects brain hemodynamics and energy metabolism. By using multi-parametric photoacoustic microscopy, the multifaceted responses of the awake mouse brain to different levels of hypercapnia are investigated. Our results show significant and vessel type-dependent increases of the vessel diameter and blood flow in response to the hypercapnic challenges, along with a decrease in oxygen extraction fraction due to elevated venous blood oxygenation. Interestingly, the increased blood flow and decreased oxygen extraction are not commensurate with each other, which leads to reduced cerebral oxygen metabolism. Further, time-lapse imaging over 2-hour chronic hypercapnic challenges reveals that the structural, functional, and metabolic changes induced by severe hypercapnia (10% CO 2 ) are not only more pronounced but more enduring than those induced by mild hypercapnia (5% CO 2 ), indicating that the extent of brain’s compensatory response to chronic hypercapnia is inversely related to the severity of the challenge. Offering quantitative, dynamic, and CO 2 level-dependent insights into the hemodynamic and metabolic responses of the brain to hypercapnia, these findings might provide useful guidance to the application of hypercapnia in brain research.