Reducing the Transformer Architecture to a Minimum

Bernhard Bermeitinger, Tomas Hrycej, Massimo Pavone, Julianus Kath, Siegfried Handschuh

2024

Abstract

Transformers are a widespread and successful model architecture, particularly in Natural Language Processing (NLP) and Computer Vision (CV). The essential innovation of this architecture is the Attention Mechanism, which solves the problem of extracting relevant context information from long sequences in NLP and realistic scenes in CV. A classical neural network component, a Multi-Layer Perceptron (MLP), complements the attention mechanism. Its necessity is frequently justified by its capability of modeling nonlinear relationships. However, the attention mechanism itself is nonlinear through its internal use of similarity measures. A possible hypothesis is that this nonlinearity is sufficient for modeling typical application problems. As the MLPs usually contain the most trainable parameters of the whole model, their omission would substantially reduce the parameter set size. Further components can also be reorganized to reduce the number of parameters. Under some conditions, query and key matrices can be collapsed into a single matrix of the same size. The same is true about value and projection matrices, which can also be omitted without eliminating the substance of the attention mechanism. Initially, the similarity measure was defined asymmetrically, with peculiar properties such as that a token is possibly dissimilar to itself. A possible symmetric definition requires only half of the parameters. All these parameter savings make sense only if the representational performance of the architecture is not significantly reduced. A comprehensive empirical proof for all important domains would be a huge task. We have laid the groundwork by testing widespread CV benchmarks: MNIST, CIFAR-10, and, with restrictions, ImageNet. The tests have shown that simplified transformer architectures (a) without MLP, (b) with collapsed matrices, and (c) symmetric similarity matrices exhibit similar performance as the original architecture, saving up to 90 % of parameters without hurting the classification performance.

Paper Citation

in Harvard Style

Bermeitinger B., Hrycej T., Pavone M., Kath J. and Handschuh S. (2024). Reducing the Transformer Architecture to a Minimum. In Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 1: KDIR; ISBN 978-989-758-716-0, SciTePress, pages 234-241. DOI: 10.5220/0012891000003838

in Bibtex Style

@conference{kdir24,
author={Bernhard Bermeitinger and Tomas Hrycej and Massimo Pavone and Julianus Kath and Siegfried Handschuh},
title={Reducing the Transformer Architecture to a Minimum},
booktitle={Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 1: KDIR},
year={2024},
pages={234-241},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012891000003838},
isbn={978-989-758-716-0},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 1: KDIR
TI - Reducing the Transformer Architecture to a Minimum
SN - 978-989-758-716-0
AU - Bermeitinger B.
AU - Hrycej T.
AU - Pavone M.
AU - Kath J.
AU - Handschuh S.
PY - 2024
SP - 234
EP - 241
DO - 10.5220/0012891000003838
PB - SciTePress