Skip to content

Observe Issue related to Elliot's Framework #24

@Faisalse

Description

@Faisalse

We are using Elliot Framework to conduct a reproducibility. However, we are facing issues with UserAutoRec, CFGAN, and LightGCN models, which produce very bad results with amazon_music, opinions and movielens_1m datasets. To handle this issue, I change the configuration file multiple times by taking random parameters and default parameters, which are given on the Elliot website for UserAutoRec, CFGAN and LightGCN models. However, I did not observe any improvement. In addition, I also analyze some articles where UserAutoRec, CFGAN, and LightGCN are picked up as baseline models to see how these models perform in other articles with amazon_music, opinions and movielens_1m datasets. Fortunately, I see their good results. I think that there is an error during the embedding of these models into the Elliot Framework,

The obtained results of UserAutoRec, CFGAN and LightGCN with amazon_music, epinions and movielens_1m datasets are given in the tables below. We also attach the used configuration files.
UserAutoRec.txt
LightGCN.txt
CFGAN.txt

image

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions