Iterative Approximate Cross-Validation
Authors: Yuetian Luo, Zhimei Ren, Rina Barber
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we illustrate the accuracy and computational efficiency of our method through a range of empirical studies. and In this section, we conduct numerical studies to investigate the empirical performance of the proposed IACV method and to verify our theoretical findings for GD, SGD, and Prox GD. |
| Researcher Affiliation | Academia | 1Data Science Institute, University of Chicago, Chicago IL 60637, USA 2Department of Statistics, University of Chicago, Chicago IL 60637, USA. |
| Pseudocode | No | The paper provides mathematical formulations and equations for the methods (e.g., (5), (6), (7), (11), (12), (13), (14)) but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code to reproduce all experiments is available at https://github.com/yuetianluo/IACV. |
| Open Datasets | No | The data is generated from a logistic regression model, with Zi = (Xi, Yi) Rp {0, 1}, with dimension p = 20 and with Xi drawn with i.i.d. N(0, 1) entries while Yi Bernoulli(exp(X i θ )/(1 + exp(X i θ ))) for true parameter vector θ which has 5 randomly chosen nonzero entries drawn as N(0, 1), and all other entries zero. The paper describes a data generation process rather than providing access to a publicly available dataset. |
| Dataset Splits | Yes | Leave-one-out CV uses n 1 out of n data points for training, and the remaining one for testing, and then repeats for each of the n data points in the sample. |
| Hardware Specification | No | The paper describes the software setup and simulation parameters (e.g., sample sizes, regularization types) but does not provide specific hardware details such as GPU/CPU models, memory, or cloud instance types used for the experiments. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions, or specific solver versions) beyond mentioning the types of models and optimization algorithms used. |
| Experiment Setup | Yes | For the gradient descent simulation, we consider sample sizes n = 250 and n = 1000, and αt = 0.5/n. and For the stochastic gradient descent simulation, we take sample size n = 1000, and test batch size K = 100 and K = 400. We choose αt based on a common strategy called epoch doubling in the literature, where we run T0 = 1000 steps with step size α = 0.5/K, then run 2T0 steps with step size α/2, and so on. |