Experience Replay for Continual Learning
Authors: David Rolnick, Arun Ahuja, Jonathan Schwarz, Timothy Lillicrap, Gregory Wayne
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show that CLEAR performs better than state-of-the-art deep learning techniques for mitigating forgetting, despite being significantly less complicated and not requiring any knowledge of the individual tasks being learned. 4.1 Catastrophic forgetting vs. interference Our first experiment (Figure 1) was designed to distinguish between two distinct concepts that are sometimes conflated, interference and catastrophic forgetting, and to emphasize the outsized role of the latter as compared to the former. |
| Researcher Affiliation | Collaboration | David Rolnick University of Pennsylvania Philadelphia, PA USA drolnick@seas.upenn.edu Deep Mind London, UK arahuja@google.com Jonathan Schwarz Deep Mind London, UK schwarzjn@google.com Timothy P. Lillicrap Deep Mind London, UK countzero@google.com Deep Mind London, UK gregwayne@google.com |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found in the paper. The method is described mathematically and textually. |
| Open Source Code | No | The paper does not provide any explicit statement about releasing source code, nor does it include links to a code repository. |
| Open Datasets | Yes | We considered a set of three distinct tasks within the DMLab set of environments [1] and reference [1]: Charles Beattie, Joel Z Leibo, Denis Teplyashin, Tom Ward, Marcus Wainwright, Heinrich Küttler, Andrew Lefrancq, Simon Green, Víctor Valdés, Amir Sadik, et al. Deep Mind Lab. Preprint ar Xiv:1612.03801, 2016. |
| Dataset Splits | No | The paper discusses training on tasks and using replay experiences but does not specify standard train/validation/test dataset splits with percentages or counts. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware used for running its experiments, such as GPU or CPU models, or cloud computing instance types. |
| Software Dependencies | No | The paper mentions algorithms like V-Trace and IMPALA and general tools like DeepMind Lab, but it does not specify software libraries or their version numbers required for replication. |
| Experiment Setup | No | The paper states that 'Our network architecture and training hyperparameters are chosen to match those in [3] and are not further optimized.' and refers to 'Further implementation details are given in Appendix A.', implying they are not fully detailed in the main text. While it mentions 'a 50-50 mixture of novel and replay experiences', this alone is not sufficient to meet the detailed hyperparameter criteria. |