Distributionally Robust Logistic Regression

Authors: Soroosh Shafieezadeh Abadeh, Peyman Mohajerin Mohajerin Esfahani, Daniel Kuhn

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate our theoretical out-of-sample guarantees through simulated and empirical experiments. 4 Numerical Results: We now showcase the power of distributionally robust logistic regression in simulated and empirical experiments.
Researcher Affiliation Academia Soroosh Shafieezadeh-Abadeh Peyman Mohajerin Esfahani Daniel Kuhn Ecole Polytechnique F ed erale de Lausanne, CH-1015 Lausanne, Switzerland {soroosh.shafiee,peyman.mohajerin,daniel.kuhn} @epfl.ch
Pseudocode No No pseudocode or clearly labeled algorithm block is present in the paper.
Open Source Code No The paper does not provide an explicit statement or link to open-source code for the methodology described.
Open Datasets Yes We validate the performance of the proposed distributionally robust logistic regression method on the MNIST dataset [30] and three popular datasets from the UCI repository: Ionosphere, Thoracic Surgery, and Breast Cancer [31]. [30] Y. Le Cun. The MNIST database of handwritten digits, 1998. http://yann.lecun.com/ exdb/mnist/. [31] K. Bache and M. Lichman. UCI machine learning repository, 2013. http://archive. ics.uci.edu/ml.
Dataset Splits No In each trial related to a UCI dataset, we randomly select 60% of data to train the models and the rest to test the performance. Similarly, in each trial related to the MNIST dataset, we randomly select 103 samples from the training dataset, and test the performance on the complete test dataset. There is no explicit mention of a separate validation set.
Hardware Specification Yes All experiments were run on an Intel XEON CPU (3.40GHz).
Software Dependencies No The paper states 'All optimization problems are implemented in MATLAB via the modeling language YALMIP [27] and solved with the state-of-the-art nonlinear programming solver IPOPT [28]' but does not provide specific version numbers for these software components.
Experiment Setup Yes For Experiment 1, 'we set κ = 1'. For Experiment 3, DRLR is implemented 'with κ = 1'. The values of ε are varied and displayed in figures, for example, 'Figure 1 displays both 1 ˆηN(ε) and the average CCR as a function of ε for different values of N'.