Inverse Covariance Estimation with Structured Groups

Authors: Shaozhe Tao, Yifan Sun, Daniel Boley

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Simulation results show significant improvement in sample complexity when the correct group structure is known. We also apply these estimators to 14,910 stock closing prices, with noticeable improvement when group sparsity is exploited.
Researcher Affiliation Collaboration Shaozhe Tao Yifan Sun Daniel Boley University of Minnesota Technicolor Research University of Minnesota taoxx120@umn.edu yifan.sun@technicolor.com boley@cs.umn.edu
Pseudocode Yes Algorithm 1 One step of Frank-Wolfe algorithm; Algorithm 2 One step of Frank-Wolfe algorithm for (4)
Open Source Code No The paper does not provide any explicit statements or links indicating the availability of open-source code for the described methodology.
Open Datasets Yes Using the Yahoo! ticker downloader [4] we downloaded 27684 tickers for different stocks. We then used the Yahoo! finance API [5] to gather daily open, high, low, close, volume, and adjusted closing prices. [4] https://pypi.python.org/pypi/ Yahoo-ticker-downloader [5] http://chart.finance.yahoo.com/table.csv?s=TICKERNAMEHERE&a=400&b=23&c=2016&d=0&e=23&f=2017&g=d&ignore=.csv
Dataset Splits Yes Define V = {1, 2, . . . , 100}, T = {101, 102, . . . , 200}, and R = {201, 202, . . . , 200 + n} as the indices of a validation, test, and train set respectively
Hardware Specification No The paper does not provide specific details on the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions tools like 'Yahoo! ticker downloader' and 'Yahoo! finance API' used for data scraping, but does not provide specific version numbers for these or any other software dependencies crucial for reproducibility.
Experiment Setup Yes To pick α and ρ, we swept powers of two in {2 5, . . . , 25} and picked the best performing ρ or α for each test. In cases where the best performing ρ or α was on the boundary, additional parameters were tested until this was no longer the case.