Controllable Invariance through Adversarial Feature Learning

Authors: Qizhe Xie, Zihang Dai, Yulun Du, Eduard Hovy, Graham Neubig

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On three benchmark tasks, namely fair and bias-free classification, language-independent generation, and lighting-independent image classification, we show that the proposed framework induces an invariant representation, and leads to better generalization evidenced by the improved performance.
Researcher Affiliation Academia Qizhe Xie, Zihang Dai, Yulun Du, Eduard Hovy, Graham Neubig Language Technologies Institute Carnegie Mellon University {qizhex, dzihang, yulund, hovy, gneubig}@cs.cmu.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Our MT code is available at https://github.com/qizhex/Controllable-Invariance
Open Datasets Yes Fair Classification For fair classification, we use three datasets... The German dataset [Frank et al., 2010]... The Adult income dataset [Frank et al., 2010]... The task of the health dataset1 is to predict whether a person will spend any days in the hospital in the following year. The sensitive variable is also the age and the dataset contains 147, 473 entries. 1www.heritagehealthprize.com, Multi-lingual Machine Translation For the multi-lingual machine translation task we use French to English (fr-en) and German to English (de-en) pairs from IWSLT 2015 dataset [Cettolo et al., 2012], Image Classification We use the Extended Yale B dataset [Georghiades et al., 2001] for our image classification task.
Dataset Splits Yes We follow the same 5-fold train/validation/test splits and feature preprocessing used in [Zemel et al., 2013, Louizos et al., 2016]. and We follow Li et al. [2014], Louizos et al. [2016] s train/test split and no validation is used
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 mentions 'Open NMT' and 'Moses multi-bleu.perl script' but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes In all of our experiments, we use Adam [Kingma and Ba, 2014] with a learning rate of 0.001., We use a batch size of 16 and the number of hidden units is set to 64. γ is set to 1 in our experiments., Every model is run for 20 epochs. γ is set to 8 and the batch size is set to 64., γ is set to 2. The batch size is set to 16 and the hidden size is set to 100.