Understanding Protein Dynamics with L1-Regularized Reversible Hidden Markov Models

Authors: Robert McGibbon, Bharath Ramsundar, Mohammad Sultan, Gert Kiss, Vijay Pande

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We implement our algorithm on GPUs and apply the method to two large protein simulation datasets generated respectively on the NCSA Bluewaters supercomputer and the Folding@Home distributed computing network.
Researcher Affiliation Academia Robert T. Mc Gibbon RMCGIBBO@STANFORD.EDU Department of Chemistry, Stanford University, Stanford CA 94305, USA Bharath Ramsundar RBHARATH@STANFORD.EDU Department of Computer Science, Stanford University, Stanford CA 94305, USA Mohammad M. Sultan MSULTAN@STANFORD.EDU Department of Chemistry, Stanford University, Stanford CA 94305, USA Gert Kiss GKISS@STANFORD.EDU Department of Chemistry, Stanford University, Stanford CA 94305, USA Vijay S. Pande PANDE@STANFORD.EDU Department of Chemistry, Stanford University, Stanford CA 94305, USA
Pseudocode No The paper describes the learning algorithm using text and equations but does not provide structured pseudocode or an algorithm block.
Open Source Code No The paper does not provide any explicit statement or link regarding the public release of its source code.
Open Datasets No The paper describes using custom-generated datasets from specific supercomputing resources and distributed networks, but does not provide concrete access information (link, DOI, repository, or explicit public availability statement) for these datasets.
Dataset Splits No The paper describes using a model selection criterion based on convergence of relaxation timescales, but does not provide specific train/test/validation dataset splits.
Hardware Specification Yes The speedup using our GPU implementation is 15 compared to our optimized CPU implementation and 75 with respect to a standard numpy implementation using K = 16 states on a NVIDIA GTX TITAN GPU / Intel Core i7 4 core Sandy Bridge CPU platform.
Software Dependencies No The paper mentions software components like 'numpy', 'Open MP', 'SSE2 intrinsics', and 'CUDA kernels' but does not provide specific version numbers for these dependencies.
Experiment Setup Yes HMMs were constructed with 2 to 6 states. We chose by monitoring the convergence of the relaxation timescales as discussed in Sec. 3.2, and set the L1 fusion penalty heuristically to a default value of λ = 0.01.