Hype-HAN: Hyperbolic Hierarchical Attention Network for Semantic Embedding

Authors: Chengkun Zhang, Junbin Gao

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Hype-HAN is applied to large scale datasets. The empirical experiments show the effectiveness of our method.
Researcher Affiliation Academia Chengkun Zhang , Junbin Gao Discipline of Business Analytics, The University of Sydney Business School The University of Sydney, Camperdown, NSW 2006, Australia {chengkun.zhang, junbin.gao}@sydney.edu.au
Pseudocode No The paper does not include pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes We test Hype-HAN on some publicly available large-scale benchmark datasets with the same protocols from [Zhang et al., 2015; Yogatama et al., 2017]. The datasets include news classification (AGnews), question/answer categorization (Yahoo Answers), sentiment analysis (Yelp and Amazon) and Wikipedia article classification (DBpedia).
Dataset Splits No Table 1 provides 'Train' and 'Test' dataset sizes (e.g., AG. 120,000 for Train and 7,600 for Test), but no explicit validation split information is provided.
Hardware Specification No The experiment described in Section 4.2 states: "This experiment is conducted without a GPU". However, for the main experiments, no specific hardware details like CPU models, GPU models, or memory are provided.
Software Dependencies No The paper mentions initializing word embeddings via 'glove-50' and training with 'manifold-aware Riemannian ADAM', but it does not specify version numbers for these or any other software dependencies.
Experiment Setup Yes We initialize the word embedding via glove-50 [Pennington et al., 2014], and we set the word/sentence hidden state as 50dimension and train the models with manifold-aware Riemannian ADAM [B ecigneul and Ganea, 2018] with learning rate 0.001. We record the prediction accuracy on the test set around epoch 10-20 on smaller datasets (AGnews, DBpedia, Yelp), and around epoch 5-10 on the large-scale datasets (Yahoo and Amazon).