Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Temporally Efficient Deep Learning with Spikes
Authors: Peter O'Connor, Efstratios Gavves, Matthias Reisser, Max Welling
ICLR 2018 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate that on MNIST, on a temporal variant of MNIST, and on Youtube-BB, a dataset with videos in the wild, our algorithm performs about as well as a standard deep network trained with backpropagation, despite only communicating discrete values between layers. |
| Researcher Affiliation | Academia | Peter O Connor, Efstratios Gavves, Matthias Reisser, Max Welling QUVA Lab University of Amsterdam Amsterdam, Netherlands EMAIL |
| Pseudocode | No | The paper provides mathematical formulas and descriptions for algorithms (e.g., in Appendix D with equations 19 and 20), but it does not present them in a structured pseudocode block format explicitly labeled as 'Pseudocode' or 'Algorithm'. |
| Open Source Code | Yes | Code is available at github.com/petered/pdnn. |
| Open Datasets | Yes | To evaluate our network s ability to learn, we train it on the standard MNIST dataset, as well as a variant we created called Temporal MNIST. [...] from the large, recently released Youtube-BB dataset Real et al. (2017). |
| Dataset Splits | No | The paper mentions 'training' and 'test' sets explicitly, as seen in Figure 5 and Appendix F, but does not explicitly describe a 'validation' set or methodology for hyperparameter tuning separate from the test set. |
| Hardware Specification | No | The paper does not specify the exact hardware components (e.g., CPU, GPU models, memory) used for running the experiments. It only references 'estimated energy-costs per op of Horowitz (2014)' for comparison without detailing their own setup. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., programming languages, libraries, frameworks with their versions) that would be needed to reproduce the experimental environment. |
| Experiment Setup | Yes | In our experiments, we choose ηk = 0.001, krel β = 0.91, kalpha = 0.91, and initialize µ0 = 1. [...] For all experiments, the PDNN started with kα = 0.5, and this was increased to kα = 0.9 after 1 epoch. |