Disentangled CVAEs with Contrastive Learning for Explainable Recommendation

Authors: Linlin Wang, Zefeng Cai, Gerard de Melo, Zhu Cao, Liang He

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate that our method generates highquality explanations and achieves new state-of-the-art results in diverse domains.
Researcher Affiliation Academia 1 East China Normal University 2 Hasso Plattner Institute, University of Potsdam 3 East China University of Science and Technology
Pseudocode No The paper describes the model architecture and training objective using mathematical equations and textual descriptions, but does not include any explicit pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any concrete statement or link regarding the availability of its source code.
Open Datasets Yes We use three large-scale datasets including Yelp1, Amazon 5-core Movie & TV2 and Trip Advisor3, and follow the common practice (Li, Zhang, and Chen 2020) to extract valid explanations and conduct pre-processing. 1www.yelp.com/dataset 2www.jmcauley.uscd.edu/data/amazon 3www.tripadvisor.com
Dataset Splits No The paper mentions 'validation loss' indicating the use of a validation set, but it does not specify the explicit percentages or sample counts for the training, validation, and test splits, nor does it provide details on the splitting methodology.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU/GPU models, memory specifications) used for running the experiments.
Software Dependencies No The paper mentions software tools like Sentence-BERT, Text Blob, and a grammar checker, along with the Adam W optimizer, but it does not provide specific version numbers for these software dependencies.
Experiment Setup Yes We set the hidden size of our 2-layer Transformer encoder and decoder to be 768, and ns, nd to be 3 for TB. The size of the fixed vocabulary is 20,000, and the batch size is 512. The hyper-parameters β and γ are consistently set to 1.0 and 0.8, respectively. For training, we use Adam W (Kingma and Ba 2015) with an initial learning rate 2 × 10−5, and decrease the learning rate by a factor of 0.8 when the decrease ratio of the validation loss is smaller than 2%.