Technical Paper on GNN-based Cryptocurrency Recommender

I'm seeking a technical paper focused on the architecture of a Recommender Engine for Cryptocurrency, employing GNN-based models such as LightGCN and GraphSage, along with a hybrid of both.

Key Components:
- A good in-depth literature review to be performed along with a detailed architectural diagram and detailed proposed architecture.
- The primary aim of this paper is to elucidate the intricacies of the model architecture and bring out their results.
- A thorough explanation of the detailed algorithmic steps alongside relevant mathematical derivations is essential.
- The paper should utilize Ethereum scanner data along with CoinGecko API data for model explanation and performance illustration.

Ideal Skills:
- Profound understanding of GNN-based models and recommender systems
- Expertise in cryptocurrency data analysis
- Strong technical writing skills
- Ability to break down complex concepts into comprehensible segments.

Please note, the paper needs to extensively compare model performance and not delve into practical application scenarios. However, this aspect may be briefly touched upon as supplementary content. Code implementation examples are not necessary, but a clear and logical sequence of model explanation along with the results and supporting images of results are necessary.
The format to be used will be provided.