Auto Encoding Neural Process for Multi-interest Recommendation
Authors: Yiheng Jiang, Yuanbo Xu, Yongjian Yang, Funing Yang, Pengyang Wang, Chaozhuo Li
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical study on 4 real world datasets demonstrates that our NP-Rec attains superior recommendation performances to several state-of-the-art baselines, where the average improvement achieves up to 13.94%. We validate the proposed algorithm on 4 real-world datasets from 2 benchmarks, and experimental results show that NP-Rec attains superior recommendation performance to several state-of-the-art SIR and MIR methods, where the average improvement achieves up to 13.94%. We verify the effectiveness and extensibility of NP model through an ablation study. |
| Researcher Affiliation | Academia | 1 Lab of Mobile Intelligent Computing, College of Computer Science and Technology, Jilin University 2 Department of Computer and Information Science, The State Key Laboratory of Internet of Things for Smart City, University of Macau 3 Beijing University of Aeronautics and Astronautics |
| Pseudocode | No | The paper describes the methodology through prose and diagrams (Figure 2) but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code https://anonymous.4open.science/r/NP-Rec-CF45 |
| Open Datasets | Yes | We conduct the validation on four widely studied datasets from two benchmarks Movie Lens (Russo et al. 2018) and Foursquare (Yang, Zhang, and Qu 2016), including ML100K, ML-1M, NYC and TKY. |
| Dataset Splits | Yes | Towards the data partition, we select each user s last previously un-interacted item as the target during recommendation procedure2, and all the prior items for training. |
| Hardware Specification | Yes | The experiments are conducted on a single server with Intel 13900K CPU and NVIDIA RTX 4090 GPU. |
| Software Dependencies | No | The paper mentions the use of 'Mamba architecture' (Gu and Dao 2023) but does not provide specific version numbers for any software libraries, frameworks, or programming languages used in the implementation. |
| Experiment Setup | Yes | Towards the sequence model, we set the dimension d = 64 and employ a layer normalization after the embedding layer along with a dropout ratio 0.3. We stack 2 Mamba layers to generate the sequence features where the internal configuration follows the original hyper-parameter settings3. As for the NP model, we employ the Re LU activation function for all MLPs to inject non-linearity, and the intermediate dimensions are set as 64. |