DropMax: Adaptive Variational Softmax
Authors: Hae Beom Lee, Juho Lee, Saehoon Kim, Eunho Yang, Sung Ju Hwang
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate our model on multiple public datasets for classification, on which it obtains significantly improved accuracy over the regular softmax classifier and other baselines. Further analysis of the learned dropout probabilities shows that our model indeed selects confusing classes more often when it performs classification. |
| Researcher Affiliation | Collaboration | KAIST1, AItrics2, South Korea, University of Oxford3, United Kingdom, {haebeom.lee, eunhoy, sjhwang82}@kaist.ac.kr juho.lee@stats.ox.ac.uk, shkim@aitrics.com |
| Pseudocode | No | The paper provides architectural diagrams (Figure 2) and mathematical formulations, but it does not include pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | Yes | The source codes are available at https://github.com/haebeom-lee/dropmax. |
| Open Datasets | Yes | 1) MNIST. This dataset [19] consists of 60, 000 images... 2) CIFAR-10. This dataset [16] consists of 10 generic object classes... 3) CIFAR-100. This dataset consists of 100 object classes... 4) AWA. This is a dataset for classifying different animal species [18]... 5) CUB-200-2011. This dataset [26] consists of 200 bird classes... |
| Dataset Splits | Yes | 1) MNIST. ...We experiment with varying number of training instances: 1K, 5K, and 55K. The validation and test set has 5K and 10K instances, respectively. ... 2) CIFAR-10. ...for each class has 5000 images for training and 1000 images for test. ... 3) CIFAR-100. ...It has 500 images for training and 100 images are for test for each class. ... 4) AWA. ...For each class, we used 50 images for test, while rest of the images are used as training set. ... 5) CUB-200-2011. ...It has 5994 training images and 5794 test images... |
| Hardware Specification | No | The paper states that experiments were implemented using “TensorFlow [1] framework” and used models like “Res Net-34” and “Res Net-18”, implying computational hardware, but it does not provide specific details on the CPU, GPU models, or other hardware specifications used for running the experiments. |
| Software Dependencies | No | The paper states “We implemented Drop Max using Tensorflow [1] framework,” but it does not specify the version of TensorFlow or any other software dependencies with version numbers. |
| Experiment Setup | Yes | The experimental setup is available in the Appendix C. |