Comply: Learning Sentences with Complex Weights inspired by Fruit Fly Olfaction
This work addresses the need for efficient and interpretable sentence embeddings in natural language processing, though it is incremental as it builds on the FlyVec model.
The authors tackled the problem of learning sentence representations by introducing Comply, a single-layer neural network with complex weights inspired by fruit fly olfaction, which outperforms FlyVec and matches larger state-of-the-art models without extra parameters.
Biologically inspired neural networks offer alternative avenues to model data distributions. FlyVec is a recent example that draws inspiration from the fruit fly's olfactory circuit to tackle the task of learning word embeddings. Surprisingly, this model performs competitively even against deep learning approaches specifically designed to encode text, and it does so with the highest degree of computational efficiency. We pose the question of whether this performance can be improved further. For this, we introduce Comply. By incorporating positional information through complex weights, we enable a single-layer neural network to learn sequence representations. Our experiments show that Comply not only supersedes FlyVec but also performs on par with significantly larger state-of-the-art models. We achieve this without additional parameters. Comply yields sparse contextual representations of sentences that can be interpreted explicitly from the neuron weights.