Single Training Dimension Selection for Word Embedding with PCA

TL;DR

This paper introduces a PCA-based method to efficiently select the optimal number of dimensions for word embeddings, reducing training effort while maintaining performance across various language tasks.

Contribution

The paper proposes a novel PCA-based approach for dimension selection in word embeddings, enabling faster training with minimal performance loss.

Findings

Achieves 10x reduction in training sets needed.

Retains optimal performance with fewer dimensions.

Applicable to multiple language tasks.

Abstract

In this paper, we present a fast and reliable method based on PCA to select the number of dimensions for word embeddings. First, we train one embedding with a generous upper bound (e.g. 1,000) of dimensions. Then we transform the embeddings using PCA and incrementally remove the lesser dimensions one at a time while recording the embeddings' performance on language tasks. Lastly, we select the number of dimensions while balancing model size and accuracy. Experiments using various datasets and language tasks demonstrate that we are able to train 10 times fewer sets of embeddings while retaining optimal performance. Researchers interested in training the best-performing embeddings for downstream tasks, such as sentiment analysis, question answering and hypernym extraction, as well as those interested in embedding compression should find the method helpful.