Inductive Quantum Embedding
Authors: Santosh Kumar Srivastava, Dinesh Khandelwal, Dhiraj Madan, Dinesh Garg, Hima Karanam, L Venkata Subramaniam
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | As an application, we show that one can achieve state-of-the-art performance on the well-known NLP task of fine-grained entity type classification by using the inductive QE approach. Our training runs 9-times faster than the original QE scheme on this task. |
| Researcher Affiliation | Industry | Santosh K. Srivastava , Dinesh Khandelwal , Dhiraj Madan , Dinesh Garg , Hima Karanam, L Venkata Subramaniam IBM Research AI, India sasriva5, dhikhand1, dmadan07, garg.dinesh, hkaranam, lvsubram@in.ibm.com |
| Pseudocode | Yes | Algorithm 1: Alternating Minimization Scheme for IQE Problem |
| Open Source Code | Yes | The code to train and evaluate our model is available at https://github.com/IBM/e2r/tree/master/neurips2020. |
| Open Datasets | Yes | We experiment with the FIGER dataset. This dataset consists of 127 different entity types arranged in two levels of the hierarchy (106 leaf and 21 internal nodes). The detailed hierarchy is shown in the supplementary material (Section 7). [16, 19] |
| Dataset Splits | Yes | We used 10% of the Test set as development set to tune parameters τ, δ, and rest 90% for the final evaluation. Table 4: Summary of FIGER dataset. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments, such as GPU or CPU models. |
| Software Dependencies | No | The paper mentions using 'PyTorch2' but does not specify a version number for this or any other software dependency. |
| Experiment Setup | Yes | Various hyper-parameter values pertaining to this step are summarized in Table 2. The last column captures the values tried during parameter selection. The second last column gives the final chosen value. The choice was made through training accuracy. ... Various hyper-parameters used in this step are given in Table 3. |