Unimodal Probability Distributions for Deep Ordinal Classification

Authors: Christopher Beckham, Christopher Pal

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate this approach in the context of deep learning on two large ordinal image datasets, obtaining promising results.
Researcher Affiliation Academia 1Montr eal Institute of Learning Algorithms, Qu ebec, Canada. Correspondence to: Christopher Beckham <christopher.beckham@polymtl.ca>.
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Code will be made available here.5 https://github.com/christopher-beckham/deep-unimodalordinal
Open Datasets Yes Diabetic retinopathy1. ... 1https://www.kaggle.com/c/diabetic-retinopathy-detection/
Dataset Splits Yes A validation set is set aside, consisting of 10% of the patients in the training set.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No The paper mentions software like Theano, Lasagne, and Keras but does not provide specific version numbers for these dependencies.
Experiment Setup Yes All experiments utilise an ℓ2 norm of 10 4, ADAM optimiser (Kingma & Ba, 2014) with initial learning rate 10 3, and batch size 128. A manual learning rate schedule is employed where we manually divide the learning rate by 10 when either the validation loss or valid set QWK plateaus (whichever plateaus last) down to a minimum of 10 4 for Adience and 10 5 for DR.