In this directory there are several examples of experiments configuration files:
basic-kge: Basic architecture with user-item and user-item-properties graph embeddings, and with BERT embeddings.hybrid-kge: Hybrid architectures (both entity-based and feature-based) with user-item and user-item-properties graph embeddings, and with BERT embeddings.basic-gnn: Basic architecture with user-item embeddings given by Graph Neural Networks.hybrid-gnn: Hybrid architecture (feature-based) with user-item embeddings given by Graph Neural Networks, and with BERT embeddings.basic-gnn-uip-1relconf: Basic architecture with user-item embeddings given by Graph Neural Networks, using the user-item-properties graph with the first relation types setting.basic-gnn-uip-2relconf: Basic architecture with user-item embeddings given by Graph Neural Networks, using the user-item-properties graph with the second relation types setting.hybrid-gnn-uip-1relconf: Hybrid architectures (both entity-based and feature-based) with user-item embeddings given by Graph Neural Networks, using the user-item-properties graph with the first relation types setting; and with BERT embeddings.hybrid-gnn-uip-2relconf: Hybrid architectures (both entity-based and feature-based) with user-item embeddings given by Graph Neural Networks, using the user-item-properties graph with the second relation types setting; and with BERT embeddings.hybrid-gnn-tweaks: Hybrid architecture (feature-based) with Attention and Residual tweaks, with user-item embeddings given by Graph Neural Networks, and with BERT embeddings.hybrid-gnn-tweaks-uip-2relconf: Hybrid architecture (feature-based) with Attention and Residual tweaks, with user-item embeddings given by Graph Neural Networks, using the user-item-properties graph with the second relation types setting; and with BERT embeddings.
Note that the all the missing parameters are assumed to be equal to the default values in config.yaml at the root
directory level.