How big can style be? Addressing high dimensionality for recommending with style
This work addresses a specific bottleneck in fashion recommendation by enabling efficient use of style information, though it is incremental as it focuses on dimensionality reduction for an existing embedding type.
The paper tackled the problem of high dimensionality in style embeddings for fashion recommender systems by proposing feature reduction methods, reducing the embedding vector from 600k to 512 features with a 99.91% memory reduction while maintaining recommendation quality.
Using embeddings as representations of products is quite commonplace in recommender systems, either by extracting the semantic embeddings of text descriptions, user sessions, collaborative relationships, or product images. In this paper, we present an approach to extract style embeddings for using in fashion recommender systems, with a special focus on style information such as textures, prints, material, etc. The main issue of using such a type of embeddings is its high dimensionality. So, we propose feature reduction solutions alongside the investigation of its influence in the overall task of recommending products of the same style based on their main image. The feature reduction we propose allows for reducing the embedding vector from 600k features to 512, leading to a memory reduction of 99.91\% without critically compromising the quality of the recommendations.