Adaptive Sampled Softmax with Kernel Based Sampling
Authors: Guy Blanc, Steffen Rendle
ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4. Experiments In this section, we empirically investigate the trade-off between bias, sampling distribution, and number of samples. 4.1. Experimental Setup |
| Researcher Affiliation | Industry | Guy Blanc 1 Steffen Rendle 2 1Work done during internship at Google, Mountain View, USA 2Google, Mountain View, USA. |
| Pseudocode | No | The paper describes algorithms in text and diagrams (Figure 1) but does not provide formal pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to open-source code for the methodology described. |
| Open Datasets | Yes | Penn Tree Bank For the NLP problem, we learn a language model on the Penn Tree Bank dataset (Marcus et al., 1999), a dataset with approximately 1 million training words and a vocabulary of size 10,000. |
| Dataset Splits | No | No specific details on train/validation/test dataset splits (percentages or counts) are provided. The paper mentions training set sizes for the YouTube datasets but not explicit splits, nor validation sets. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper mentions using a TensorFlow-based LSTM implementation indirectly via a URL, but no specific software versions (e.g., TensorFlow version, Python version, library versions) are provided. |
| Experiment Setup | No | The paper describes some model architecture details (e.g., LSTM units per layer changed from 650 to 200) and sampling distributions, but it does not provide specific experimental setup details such as learning rates, batch sizes, optimizer settings, or other hyperparameters. |