When Optimizing $f$-Divergence is Robust with Label Noise
Authors: Jiaheng Wei, Yang Liu
ICLR 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we validate our analysis of Df measures robustness via a set of empirical evaluations on 5 datasets: MNIST (Le Cun et al. (1998)), Fashion-MNIST (Xiao et al. (2017)), CIFAR-10 and CIFAR-100 (Krizhevsky et al. (2009)), and Clothing1M (Xiao et al. (2015)). |
| Researcher Affiliation | Academia | Jiaheng Wei and Yang Liu Department of Computer Science and Engineering University of California, Santa Cruz Santa Cruz, CA 95060, USA {jiahengwei, yangliu}@ucsc.edu |
| Pseudocode | Yes | Algorithm 1 Maximizing Df measures: one step gradient |
| Open Source Code | Yes | Our code is available at https: //github.com/UCSC-REAL/Robust-f-divergence-measures. |
| Open Datasets | Yes | In this section, we validate our analysis of Df measures robustness via a set of empirical evaluations on 5 datasets: MNIST (Le Cun et al. (1998)), Fashion-MNIST (Xiao et al. (2017)), CIFAR-10 and CIFAR-100 (Krizhevsky et al. (2009)), and Clothing1M (Xiao et al. (2015)). |
| Dataset Splits | No | The paper mentions training and testing datasets, but does not provide explicit details on the percentage splits or methodology for creating these splits within the provided text. It states 'Omitted experiment details are available in the appendix.'. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers (e.g., programming languages, libraries, frameworks). |
| Experiment Setup | Yes | In experiments, since the estimation of product noisy distribution are unstable when trained on CIFAR-100 training dataset, we use CE as a warm-up (120 epochs) and then switch to train with Df measures. ... At step t, update ht by ascending its stochastic gradient with learning rate ηt |