Learning to Model the Tail
Authors: Yu-Xiong Wang, Deva Ramanan, Martial Hebert
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate results on image classification datasets (SUN, Places, and Image Net) tuned for the long-tailed setting, that significantly outperform common heuristics, such as data resampling or reweighting. |
| Researcher Affiliation | Academia | Yu-Xiong Wang Deva Ramanan Martial Hebert Robotics Institute, Carnegie Mellon University {yuxiongw,dramanan,hebert}@cs.cmu.edu |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement or link for open-source code for the described methodology. |
| Open Datasets | Yes | We demonstrate results on image classification datasets (SUN, Places, and Image Net) tuned for the long-tailed setting, that significantly outperform common heuristics, such as data resampling or reweighting. [SUN-397 dataset [14], Places-205 [7], ILSVRC 2012 classification dataset [5]] |
| Dataset Splits | Yes | Following the experimental setup in [61, 62, 63], we first randomly split the dataset into train, validation, and test parts using 50%, 10%, and 40% of the data, respectively. |
| Hardware Specification | No | The paper mentions "NVIDIA for donating GPUs and AWS Cloud Credits for Research program" but does not specify exact GPU models, specific AWS instance types, or other detailed hardware specifications used for running experiments. |
| Software Dependencies | No | The paper mentions using "Caffe [59]" but does not specify a version number for Caffe or any other software dependencies. |
| Experiment Setup | Yes | We use 0.01 as the negative slope for leaky Re LU. Computation is naturally divided into two stages... fine-tuning is performed for around 60 epochs using SGD with an initial learning rate of 0.01, which is reduced by a factor of 10 around every 30 epochs. |