Online and Stochastic Gradient Methods for Non-decomposable Loss Functions

Authors: Purushottam Kar, Harikrishna Narasimhan, Prateek Jain

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

Reproducibility Variable Result LLM Response
Research Type Experimental We then use extensive experimentation on real life and benchmark datasets to establish that our method can be orders of magnitude faster than a recently proposed cutting plane method.
Researcher Affiliation Collaboration Microsoft Research, INDIA Indian Institute of Science, Bangalore, INDIA
Pseudocode Yes Algorithm 1 1PMB: Single-Pass with Mini-batches
Open Source Code No The paper does not provide any explicit statement about releasing source code for the described methodology or a link to a code repository.
Open Datasets Yes We used several data sets for our experiments [...] and the remaining datasets are taken from the UCI repository [22]. [...] [22] A. Frank and Arthur Asuncion. The UCI Machine Learning Repository. http://archive.ics.uci.edu/ml, 2010.
Dataset Splits Yes We used 70% of the data set for training and the remaining for testing, with the results averaged over 5 random train-test splits. Tunable parameters such as step length scale were chosen using a small validation set.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., 'Python 3.8, PyTorch 1.9') required to replicate the experiment.
Experiment Setup Yes The epoch lengths/buffer sizes were set to 500 in all experiments.