Practical Gauss-Newton Optimisation for Deep Learning

Authors: Aleksandar Botev, Hippolyt Ritter, David Barber

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

Reproducibility Variable Result LLM Response
Research Type Experimental We performed experiments training deep autoencoders on three standard grey-scale image datasets and classifying hand-written digits as odd or even.
Researcher Affiliation Academia 1University College London, London, United Kingdom 2Alan Turing Institute, London, United Kingdom.
Pseudocode No The paper contains mathematical equations and derivations but does not include any structured pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing source code for their methodology, nor does it provide a direct link to a code repository.
Open Datasets Yes MNIST consists of 60, 000 28 28 images of hand-written digits.
Dataset Splits Yes MNIST consists of 60, 000 28 28 images of hand-written digits. We used only the first 50, 000 images for training (since the remaining 10, 000 are usually used for validation).
Hardware Specification Yes Experiments were run on a workstation with a Titan Xp GPU and an Intel Xeon CPU E5-2620 v4 @ 2.10GHz.
Software Dependencies No The paper mentions that methods were implemented 'using Theano (Theano Development Team, 2016) and Lasagne (Dieleman et al., 2015)', but it does not provide specific version numbers for these software dependencies in the text.
Experiment Setup Yes The layer sizes are D-1000-500-250-30-250-500-1000-D, where D is the dimensionality of the input.