Better Fine-Tuning via Instance Weighting for Text Classification

Authors: Zhi Wang, Wei Bi, Yan Wang, Xiaojiang Liu7241-7248

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that IW-Fit can consistently improve the classification accuracy on the target domain.
Researcher Affiliation Collaboration 1Department of Control and Systems Engineering, Nanjing University, Nanjing, China 2Tencent AI Lab, Shenzhen, China
Pseudocode Yes Algorithm 1: Instance Weighting based Fine-tuning (IW-Fit) for text classification
Open Source Code No The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes The source dataset from 2015 Amazon reviews (Zhang, Zhao, and Le Cun, 2015; Conneau et al., 2016) is mainly about products of online shopping. The target dataset of Yelp reviews is obtained from the latest 2018 Yelp Dataset Challenge
Dataset Splits No The paper specifies 10,000 training instances and 1,000 testing instances for the target dataset but does not explicitly mention a separate validation split or its size.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types, memory) used to run the experiments.
Software Dependencies No The paper mentions software components like LSTM and CNN but does not provide specific version numbers for any software libraries, frameworks, or programming languages used.
Experiment Setup Yes The hyperparameters are set as: vocabulary size = 30, 000, embedding size = 100, LSTM hidden size = 200, FC layer size = 200, dropout rate = 0.2, L2 regularization = 1e 4, batch size = 32, fine-tuning epochs = 50, burn-in epochs = 10, learning rate = 1e 3.