Understanding Synthetic Gradients and Decoupled Neural Interfaces

Authors: Wojciech Marian Czarnecki, Grzegorz Świrszcz, Max Jaderberg, Simon Osindero, Oriol Vinyals, Koray Kavukcuoglu

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct an empirical analysis of the learning dynamics on easily analysable artificial data. We create 2 and 100 dimensional versions of four basic datasets (details in the Supplementary Materials Section D) and train four simple models (a linear model and a deep linear one with 10 hidden layers, trained to minimise MSE and log loss) with regular backprop and with a SG-based alternative to see whether it (numerically) converges to the same solution.
Researcher Affiliation Industry 1Deep Mind, London, United Kingdom.
Pseudocode No The paper does not contain any sections or figures explicitly labeled as 'Pseudocode' or 'Algorithm', nor does it present structured steps formatted like code.
Open Source Code No The paper does not contain any explicit statement about releasing source code or provide a link to a code repository for the methodology described.
Open Datasets Yes We train deep relu networks of varied depth (up to 50 hidden layers) with batch-normalisation and with two different activation functions on MNIST and compare models trained with full backpropagation to variants that employ a SG module in the middle of the hidden stack.
Dataset Splits No The paper mentions experiments on MNIST, but it does not provide specific details on dataset splits (e.g., percentages, sample counts, or citations to standard splits) for training, validation, or testing.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments, only general statements about 'training neural networks'.
Software Dependencies No The paper does not provide specific details on software dependencies with version numbers (e.g., programming languages, libraries, or frameworks with their respective versions) used for the experiments.
Experiment Setup Yes We train deep relu networks of varied depth (up to 50 hidden layers) with batch-normalisation and with two different activation functions on MNIST and compare models trained with full backpropagation to variants that employ a SG module in the middle of the hidden stack. ... We train with a small L2 penalty added to weights...