demoleiwang / EC4SRec

Code for the CIKM2022 Paper "Explanation Guided Contrastive Learning for Sequential Recommendation"

EC4SRec (Updating)

Code for the CIKM2022 Paper "Explanation Guided Contrastive Learning for Sequential Recommendation"

EC4SRec News

09/27/2022: We updated DuoRec on ml-100k.
09/19/2022: We updated CL4SRec w/ X-Aug (Ours) on ml-100k.
09/17/2022: We updated SASRec and CL4SRec on ml-100k.

EC4SRec To-Do list

• Update SASRec.
• Update CL4SRec.
• Update CL4SRec w/ X-Aug (Ours).
• Update CoSeRec.
• Update CoSeRec w/ X-Aug (Ours).
• Update ContraRec.
• Update ContraRec w/ X-Aug (Ours).
• Update DuoRec.
• Update DuoRec w/ X-Aug (Ours).
• Update ICL.
• Update EC4SRec (Ours).

Quick Start

Command for CL4SRec

python run_train.py --dataset=ml-100k --model=SASRec --method=CL4SRec --gpu_id=13
python run_train.py --dataset=Amazon_Beauty --model=SASRec --method=CL4SRec --gpu_id=4

Command for CL4SRec w/ X-Aug (Ours)

python run_train.py --dataset=ml-100k --model=SASRec --method=CL4SRec_XAUG --xai_method=occlusion --gpu_id=2
python run_train.py --dataset=Amazon_Beauty --model=SASRec --method=CL4SRec_XAUG --xai_method=occlusion --gpu_id=12

Command for DuoRec

python run_train.py --dataset=ml-100k --model=SASRec --method=DuoRec --gpu_id=2
python run_train.py --dataset=Amazon_Beauty --model=SASRec --method=DuoRec --gpu_id=12

Command for DuoRec w/ X-Aug (Ours)

python run_train.py --dataset=ml-100k --model=SASRec --method=DuoRec_XAUG --xai_method=occlusion --gpu_id=2
python run_train.py --dataset=Amazon_Beauty --model=SASRec --method=DuoRec_XAUG --xai_method=occlusion --gpu_id=12

Performance

5-core Setting

Experimtents on ML-100K.

model	Recall@5	Recall@10	Recall@20	NDCG@5	NDCG@10	NDCG@20
CL4SRec	0.0753	0.1273	0.2216	0.0452	0.0617	0.0856
+X-Aug (Occlusion)	0.0806	0.1389	0.2418	0.0454	0.0642	0.0899
CoSeRec
+X-Aug (Occlusion)
ContraRec
+X-Aug (Occlusion)
DuoRec	0.0764	0.1198	0.2216	0.0462	0.0602	0.0860
+X-Aug (Occlusion)	0.0753	0.1315	0.2238	0.0456	0.0633	0.0866
ICL
EC4SRec (Occlusion)

Experimtents on Amazon_Beauty

model	Recall@5	Recall@10	Recall@20	NDCG@5	NDCG@10	NDCG@20
CL4SRec
+X-Aug (Occlusion)
CoSeRec
+X-Aug (Occlusion)
ContraRec
+X-Aug (Occlusion)
DuoRec
+X-Aug (Occlusion)
ICL
EC4SRec (Occlusion)

Citations

@article{wang2022explanation,
  title={Explanation Guided Contrastive Learning for Sequential Recommendation},
  author={Wang, Lei and Lim, Ee-Peng and Liu, Zhiwei and Zhao, Tianxiang},
  journal={arXiv preprint arXiv:2209.01347},
  year={2022}
}