Description: We propose VecKM, a highly efficient local point cloud encoder. VecKM is the first successful attempt to reduce the computation and memory costs from O(n^2+nKd) to O(nd) by sacrificing a marginal factor, where n is the size of the point cloud and K is neighborhood size. The efficiency is primarily due to VecKM's unique factorizable property that eliminates the need of explicitly grouping points into neighborhoods.

Description: We introduce Hyper-Dimensional Function Encoding (HDFE), which generates a fixed-length vector representation of continuous objects, such as functions, invariant to sample distribution and density. This invariance allows HDFE to consistently encode continuous objects for neural networks without the need for training. It maps objects into an organized embedding space, aiding downstream task training. HDFE is also decodable, enabling neural networks to regress continuous objects via their encodings. Thus, HDFE acts as an interface for processing continuous objects in machine learning tasks like classification and regression.

Description: We introduce a novel approach using binary hypervectors to encode neural network output signals for collaboration. These hypervectors are efficiently combined through consensus summation. This method outperforms or matches state-of-the-art techniques with minimal computational overhead. In addition, this consensus-based learning process can adapt in real-time, accommodating the addition or removal of models.

Description: We reveal the negative health impacts of PM2.5 pollution and its connection to urbanization and long-term meteorological changes. We reveal that urbanization tends to increase global PM2.5 levels and associated mortality, with urban residents being more vulnerable to adverse meteorological conditions. Machine learning analysis shows that meteorological influences on PM2.5-related health burdens have grown over time, emphasizing the need for climate mitigation efforts to address this issue.

Description: We use an ensemble neural network to estimate the long-term wet deposition of harmful substances (nss-SO42– and NO3–) resulting from human emissions over East Asia. The model's accuracy is high, with R2 values of 0.90 for nss-SO42– and 0.85 for NO3–. We identify hotspots for these depositions in southwestern, central, and eastern China. Notably, we find that acidity in rainwater is shifting from sulfur to nitrogen dominance in most regions. Additionally, We reveal that some ecosystems, particularly freshwater bodies and grasslands, are experiencing wet depositions that exceed critical load thresholds, indicating a need for stricter, collaborative emission control measures in East Asia to protect these ecosystems.

Description: We address the importance of understanding secondary inorganic aerosols in PM2.5 pollution. We use an ensemble neural network model combining numerical models and data to estimate aerosol concentrations in the Greater Bay Area in 2005 and 2015. The model outperforms traditional air quality models, showing decreasing concentrations of SO2-4, NO-3, and NH+4 from 2005 to 2015. Analysis reveals reduced sulfate formation due to emission controls, while increased ammonia availability leads to more sulfate neutralization. We underscore the need for controlling ammonia emissions to further reduce aerosol concentrations in the future.