DRUM: End-To-End Differentiable Rule Mining On Knowledge Graphs
Authors: Ali Sadeghian, Mohammadreza Armandpour, Patrick Ding, Daisy Zhe Wang
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section we evaluate DRUM on statistical relation learning and knowledge base completion. We also empirically assess the quality and interpretability of the learned rules. We implement our method in TensorFlow [1] and train on Tesla K40 GPUs. We use ADAM [19] with learning rate and batch size of 0.001 and 64, respectively. We set both the hidden state dimension and head relation vector size to 128. We did gradient clipping for training the RNNs and used LSTMs [17] for both directions. fθ is a single layer fully connected. We followed the convention in the existing literature [41] of splitting the data into three categories of facts, train, and test. The code and the datasets for all the experiments will be publicly available. |
| Researcher Affiliation | Academia | 1 Department of Computer Science, University of Florida 2 Department of Statistics, Texas A&M University |
| Pseudocode | No | The paper does not contain any pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code and the datasets for all the experiments will be publicly available. |
| Open Datasets | Yes | Datasets: Our experiments were conducted on three different datasets [20]. The Unified Medical Language System (UMLS) consists of biomedical concepts such as drug and disease names and relations between them such as diagnosis and treatment. Kinship contains kinship relationships among members of a Central Australian native tribe. The Family data set contains the bloodline relationships between individuals of multiple families. Statistics about each data set are shown in Table 1. ... We evaluate our proposed model in inductive and transductive link prediction tasks on two widely used knowledge graphs Word Net [18, 24] and Freebase [3]. |
| Dataset Splits | Yes | We followed the convention in the existing literature [41] of splitting the data into three categories of facts, train, and test. ... To train the model, we split the training file into facts and new training file with the ratio of three to one. ... Table 3: Datasets statistics for Knowledge base completion. #Train 86,835 #Valid 3,034 #Test 3,134 |
| Hardware Specification | Yes | We implement our method in TensorFlow [1] and train on Tesla K40 GPUs. |
| Software Dependencies | No | The paper mentions TensorFlow [1] but does not provide a specific version number. It mentions ADAM [19] and LSTMs [17] as methods used, not software dependencies with versions. |
| Experiment Setup | Yes | We use ADAM [19] with learning rate and batch size of 0.001 and 64, respectively. We set both the hidden state dimension and head relation vector size to 128. We did gradient clipping for training the RNNs and used LSTMs [17] for both directions. fθ is a single layer fully connected. |