skip to main content
Explore Scholarly Publications and Datasets in the NSF-PAR

  • Home
  • Search Results
  • Page 1 of 36

Search for: All records

Creators/Authors contains: "Lee, K."

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. ; ; ( , International Conference on Learning Representations)
    Free, publicly-accessible full text available January 1, 2023
  2. ; ; ( , The Journal of Physical Chemistry B)
    Full Text Available 
  3. ; ; ; ; ( , Fourth Conference on Machine Learning and Systems)
    Distributed model training suffers from communication bottlenecks due to frequent model updates transmitted across compute nodes. To alleviate these bottlenecks, practitioners use gradient compression techniques like sparsification, quantization, or low-rank updates. The techniques usually require choosing a static compression ratio, often requiring users to balance the trade-off between model accuracy and per-iteration speedup. In this work, we show that such performance degradation due to choosing a high compression ratio is not fundamental. An adaptive compression strategy can reduce communication while maintaining final test accuracy. Inspired by recent findings on critical learning regimes, in which small gradient errors can have irrecoverablemore »impact on model performance, we propose Accordion a simple yet effective adaptive compression algorithm. While Accordion maintains a high enough compression rate on average, it avoids over-compressing gradients whenever in critical learning regimes, detected by a simple gradient-norm based criterion. Our extensive experimental study over a number of machine learning tasks in distributed environments indicates that Accordion, maintains similar model accuracy to uncompressed training, yet achieves up to 5.5x better compression and up to 4.1x end-to-end speedup over static approaches. We show that Accordion also works for adjusting the batch size, another popular strategy for alleviating communication bottlenecks.« less
    Accepted Manuscript Full Text Available
  4. ; ; ; ( , The Astrophysical Journal)
    Full Text Available 
  5. ; ; ; ( , Neural Information Processing Systems (NeurIPS))
    Accepted Manuscript Full Text Available
  6. ; ; ; ; ; ; ( , Environmental Science: Water Research & Technology)
    Antimicrobial resistance is a well-documented public health concern. The role that drinking water distribution pipes have as sources of antibiotic resistance genes (ARGs) is not well known. Metals are a known stressor for antibiotic resistance development, implying that aging metal-pipe infrastructure could be a source of ARGs. The objective of this study was to determine if ARGs, metal resistance genes (MRGs), and intI 1 were pervasive across various pipe biofilm sample types (biomass surfaces, pipe surfaces, corrosion tubercles, and under corrosion tubercles) and if the resistance genes associated with particular microbial taxa. Eight sample types in triplicate ( n =more »24) were taken from inside a >100 year-old, six ft. section of a full-scale chloraminated cast iron drinking water main. Droplet digital PCR (ddPCR) was employed as a novel approach to quantify ARGs in pipes from full-scale drinking water distribution systems (DWDS) because it yielded higher detection frequencies than quantitative PCR (qPCR). Illumina sequencing was employed to characterize the microbial community based on 16S rRNA genes. ARGs and MRGs were detected in all 24 pipe samples. Every sample contained targeted genes. Interestingly, the mean absolute abundances of ARGs and MRGs only varied by approximately one log value across sample types, but the mean relative abundances (copy numbers normalized to 16S rRNA genes) varied by over two log values. The ARG and MRGs concentrations were not significantly different between sample types, despite significant changes in dominant microbial taxa. The most abundant genera observed in the biofilm communities were Mycobacterium (0.2–70%), and β-lactam resistance genes bla TEM , bla SHV , and the integrase gene of class 1 integrons ( intI 1) were positively correlated with Mycobacterium . The detection of ARGs, MRGs, and class 1 integrons across all sample types within the pipe indicates that pipes themselves can serve as sources for ARGs in DWDS. Consequently, future work should investigate the role of pipe materials as well as corrosion inhibitors to determine how engineering decisions can mitigate ARGs in drinking water that stem from pipe materials.« less
    Full Text Available 
  7. ; ; ; ; ; ; ; ; ; ( , The Astrophysical Journal)
    Full Text Available 
  8. ; ; ; ; ; ; ; ; ; ; et al ( , Nature Materials)
    Free, publicly-accessible full text available November 1, 2022
  9. ; ; ; ; ; ; ; ; ; ; et al ( , Physical Review C)
    Full Text Available 
  10. ; ; ; ; ; ; ; ; ; ; et al ( , International Conference on Medical Image Computing and Computer Assisted Interventions (MICCAI))
    Electron microscopy (EM) enables the reconstruction of neural circuits at the level of individual synapses, which has been transformative for scientific discoveries. However, due to the complex morphology, an accurate reconstruction of cortical axons has become a major challenge. Worse still, there is no publicly available large-scale EM dataset from the cortex that provides dense ground truth segmentation for axons, making it difficult to develop and evaluate large-scale axon reconstruction methods. To address this, we introduce the AxonEM dataset, which consists of two 30x30x30 cubic mm EM image volumes from the human and mouse cortex, respectively. We thoroughly proofread overmore »18,000 axon instances to provide dense 3D axon instance segmentation, enabling large- scale evaluation of axon reconstruction methods. In addition, we densely annotate nine ground truth subvolumes for training, per each data volume. With this, we reproduce two published state-of-the-art methods and provide their evaluation results as a baseline. We publicly release our code and data at https://connectomics-bazaar.github.io/proj/ AxonEM/index.html to foster the development of advanced methods.« less
    Full Text Available