Online Portfolio Selection with Group Sparsity

Authors: Puja Das, Nicholas Johnson, Arindam Banerjee

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on two real-world datasets and use the Global Industry Classification Standard to group the stocks into sectors for 22 years of the benchmark NYSE dataset with 30 stocks and 8 sectors and 22 years of a S&P500 dataset with 243 stocks and 9 sectors. Our experiments show that our sparse group lazy portfolios can take advantage of the sector information to beat the market and are scalable with transaction costs.
Researcher Affiliation Academia Puja Das and Nicholas Johnson and Arindam Banerjee Department of Computer Science and Engineering University of Minnesota Minneapolis, MN 55455 {pdas, njohnson, banerjee}@cs.umn.edu
Pseudocode Yes Algorithm 1 OLU-GS Algorithm with ADMM; Algorithm 2 Portfolio Selection with Group Sparsity
Open Source Code No The paper does not provide any statement or link indicating that the source code for the described methodology is open-source or publicly available.
Open Datasets Yes The NYSE dataset (Helmbold et al. 1998; Agarwal et al. 2006; Borodin, El-Yaniv, and Gogan 2004; Cover 1991) consists of 36 stocks with data accumulated over a period of 22 years from July 3, 1962 to December 31, 1984. The S&P500 dataset consists of 258 stocks with data accumulated over a period of 22 years from 1991 to 2012.
Dataset Splits No The paper mentions using the NYSE and S&P500 datasets but does not specify explicit training, validation, or test splits, percentages, or sample counts for these datasets in the experimental setup.
Hardware Specification No The paper does not specify any hardware details (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers.
Experiment Setup Yes The parameters consist of λ1: weight on group sparsity norm, λ2: lazy updates weight, η: weight on the ℓ2 norm, and β: the parameter for the augmentation term. For all our experiments, we set β = 2 which we found to give reasonable accuracy and use group lasso for group sparsity. [...] We experimented extensively with a large range of λ1, λ2, and η values from 1e 9 to 1 to observe their effect on group sparsity and lazy updates to our portfolio.