Discriminative Gaifman Models
Authors: Mathias Niepert
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments The aim of the experiments is to understand the efficiency and effectiveness of Gaifman models for typical knowledge base inference problems. We evaluate the proposed class of models with two data sets derived from the knowledge bases WORDNET and FREEBASE [2]. Both data sets consist of a list of statements r(d1, d2) that are known to be true. |
| Researcher Affiliation | Industry | Mathias Niepert NEC Labs Europe Heidelberg, Germany mathias.niepert@neclabs.eu |
| Pseudocode | Yes | Algorithm 1 GENNEIGHS: Computes a list of w neighborhoods of size k for an input tuple d. |
| Open Source Code | No | The paper does not provide any statement or link regarding the public availability of its source code. |
| Open Datasets | Yes | We evaluate the proposed class of models with two data sets derived from the knowledge bases WORDNET and FREEBASE [2]. Both data sets consist of a list of statements r(d1, d2) that are known to be true. For a detailed description of the data sets, whose statistics are listed in Table 1, we refer the reader to previous work [4]. |
| Dataset Splits | No | Table 1 lists '# train' and '# test' samples for the WN18 and FB15k datasets but does not explicitly specify the proportions or sizes of training, validation, and test splits, nor does it provide details on how the data was partitioned for reproducibility. |
| Hardware Specification | Yes | All experiments were run on commodity hardware with 64G RAM and a single 2.8 GHz CPU. |
| Software Dependencies | No | The paper mentions using a neural network architecture but does not specify any software libraries, frameworks, or their version numbers (e.g., Python, PyTorch, TensorFlow, CUDA). |
| Experiment Setup | Yes | We use a neural network architecture with two hidden layers, each having 100 units and sigmoid activations, dropout of 0.2 on the input layer, and a softmax layer. We trained one model per relation and left the hyper-parameters fixed across models. To compute the probabilities, we averaged the probabilities of N = 1, 2, or 3 generated (r, k)-neighborhoods. |