Search for: All records

Masked Autoencoder (MAE) is a notable method for self-supervised pretraining in visual representation learning. It operates by randomly masking image patches and reconstructing these masked patches using the unmasked ones. A key limitation of MAE lies in its disregard for the varying informativeness of different patches, as it uniformly selects patches to mask. To overcome this, some approaches propose masking based on patch informativeness. However, these methods often do not consider the specific requirements of downstream tasks, potentially leading to suboptimal representations for these tasks. In response, we introduce the Multi-level Optimized Mask Autoencoder (MLO-MAE), a novel framework that leverages end-to-end feedback from downstream tasks to learn an optimal masking strategy during pretraining. Our experimental findings highlight MLO-MAE's significant advancements in visual representation learning. Compared to existing methods, it demonstrates remarkable improvements across diverse datasets and tasks, showcasing its adaptability and efficiency. Our code is available at https://github.com/Alexiland/MLO-MAE 

Recent drought, wildfires and rising temperatures in the western US highlight the urgency of increasing resiliency in overstocked forests. However, limited valuation information hinders the broader participation of beneficiaries in forest management. We assessed how historical disturbances in California's Central Sierra Nevada affected live biomass, forest water use and carbon uptake and estimated marginal values of these changes. On average, low‐severity wildfire caused greater declines in forest evapotranspiration (ET), gross primary productivity (GPP) and live biomass than did commercial thinning. Low‐severity wildfires represent proxies for prescribed burns and both function as biomass removal to alleviate overstocked conditions. Increases in potential runoff over 15 years post‐disturbance were valued at $108,000/km²for commercial thinning versus $234,000/km²for low‐severity wildfire, based on historical water prices. Respective declines in GPP were valued at −$305,000 and −$1,317,000/km², based on an average social cost of carbon. Considering biomass levels created by commercial thinning and low‐severity fire as more‐sustainable management baselines for overstocked forests, carbon uptake over 15 years post‐disturbance can be viewed as a benefit rather than loss. Realizing this benefit upon management re‐entry may require sequestering thinned material. High‐severity wildfire and clearcutting resulted in greater declines in ET and thus greater potential water benefits but also substantial declines in GPP and live carbon. These lessons from historical disturbances indicate what benefit ranges from fuels treatments can be expected from more‐sustainable management of mixed‐conifer forests and the importance of setting an appropriate baseline. 

Abstract Alzheimer's disease is the leading cause of dementia. The long progression period in Alzheimer's disease provides a possibility for patients to get early treatment by having routine screenings. However, current clinical diagnostic imaging tools do not meet the specific requirements for screening procedures due to high cost and limited availability. In this work, we took the initiative to evaluate the retina, especially the retinal vasculature, as an alternative for conducting screenings for dementia patients caused by Alzheimer's disease. Highly modular machine learning techniques were employed throughout the whole pipeline. Utilizing data from the UK Biobank, the pipeline achieved an average classification accuracy of 82.44%. Besides the high classification accuracy, we also added a saliency analysis to strengthen this pipeline's interpretability. The saliency analysis indicated that within retinal images, small vessels carry more information for diagnosing Alzheimer's diseases, which aligns with related studies.