giacoballoccu / explanation-quality-recsys

Post Processing Explanations Paths in Path Reasoning Recommender Systems with Knowledge Graphs

Hi there 👋

I'm a PhD student at University of Cagliari (UniCa) and my research interest lay in Recommender Systems and end-user Explainability tecniques. I'm currently working on designing end-to-end Language Models for Explainable Recommendation. Previously I mostly worked with Knowledge-Aware Recommender Systems and Path Reasoning methods.

I'm teaching assistant of Algorithm and Data Structure in my University and I previously interned for 4 months at Amazon as Applied Scientist in the Targeting Science Ads team.

I train my models using Paperspace which gives you the chance to lease paying monthly subscription GPUs such as A100 80Gb, A6000 etc. Click here to get my referral and get 10$ of credit for free.

• 🎓 Google Scholar
• 👨‍💻 LeetCode
• 📫 Reach me out

.

explanation-quality-recsys's Issues

The chosen optimization doesn't exist...

run

python main.py --dataset=ml1m --opt=soft--metrics=LIR

i get this problem

The chosen optimization doesn't exist...

Traceback (most recent call last): File "main.py", line 208, in <module> soft_optimization_ETD(path_data) File "E:\FengHanWen\explanation-quality-recsys-main\optimizations.py", line 42, in soft_optimization_ETD current_item_pred_paths = pred_path[uid][pid] KeyError: 919

i have got the three datasets,
the lastfm and ml1m in preprocessed_datasets.tar.gz put in explanation-quality-recsys-main/datasets
the lastfm and ml1m in lastfmpaths ,ml1mpaths put in explanation-quality-recsys-main/paths
the lastfm and ml1m in lastfmtmp,ml1mtmp put in explanation-quality-recsys-main/models/PGPR/tmp

after running python main.py --dataset=lastfm --opt=softETD
i get the error
Traceback (most recent call last):
File "main.py", line 208, in
soft_optimization_ETD(path_data)
File "E:\FengHanWen\explanation-quality-recsys-main\optimizations.py", line 42, in soft_optimization_ETD
current_item_pred_paths = pred_path[uid][pid]
KeyError: 919

i change command --opt=softSEP/softLIR, but there is the same problem as above
i can't get results

ABOUT PGPR baseline

When executing the adapted PGPR baseline, I had some problems;
I tried to use python main.py --dataset=ml1m --opt=soft --metrics=LIR/s_to_opt. and some other changes . However, they were all wrong
The wrong wae showen as follows:
main.py [-h] [--dataset DATASET] [--agent_topk AGENT_TOPK] [--opt OPT] [--alpha ALPHA] [--eval_baseline EVAL_BASELINE] [--log_enabled LOG_ENABLED] [--save_baseline_rec_quality_avgs SAVE_BASELINE_REC_QUALITY_AVGS]
[--save_baseline_exp_quality_avgs SAVE_BASELINE_EXP_QUALITY_AVGS] [--save_baseline_rec_quality_distributions SAVE_BASELINE_REC_QUALITY_DISTRIBUTIONS]
[--save_baseline_exp_quality_distributions SAVE_BASELINE_EXP_QUALITY_DISTRIBUTIONS] [--save_after_rec_quality_avgs SAVE_AFTER_REC_QUALITY_AVGS] [--save_after_exp_quality_avgs SAVE_AFTER_EXP_QUALITY_AVGS]
[--save_after_rec_quality_distributions SAVE_AFTER_REC_QUALITY_DISTRIBUTIONS] [--save_after_exp_quality_distributions SAVE_AFTER_EXP_QUALITY_DISTRIBUTIONS] [--save_overall SAVE_OVERALL]
main.py: error: unrecognized arguments: --metrics=SEP/s_to_opt