Overcoming Catastrophic Forgetting by Incremental Moment Matching

Authors: Sang-Woo Lee, Jin-Hwa Kim, Jaehyun Jun, Jung-Woo Ha, Byoung-Tak Zhang

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

Reproducibility Variable Result LLM Response
Research Type Experimental The experimental results show that IMM achieves state-of-the-art performance by balancing the information between an old and a new network.
Researcher Affiliation Collaboration Sang-Woo Lee1, Jin-Hwa Kim1, Jaehyun Jun1, Jung-Woo Ha2, and Byoung-Tak Zhang1,3 Seoul National University1 Clova AI Research, NAVER Corp2 Surromind Robotics3
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes The source code for the experiments is available in Github repository1. 1https://github.com/btjhjeon/IMM_tensorflow
Open Datasets Yes We analyze our approach on a variety of datasets including the MNIST, CIFAR-10, Caltech-UCSDBirds, and Lifelog datasets.
Dataset Splits No Consider that tuned hyperparameter setting is often used in previous works of continual learning as it is difficult to define a validation set in their setting.
Hardware Specification No The paper does not provide specific hardware details used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes Hyperparam denotes the main hyperparameter of each algorithm. For IMM with transfer, only α is tuned. The numbers in the parentheses refer to standard deviation. Every IMM uses weight-transfer. Table 1 lists specific hyperparameter values like 'λ in (10)', 'p in (11)', 'α2 in (4)', 'α2 in (7)' for both untuned and tuned settings.