Measuring Representational Shifts in Continual Learning: A Linear Transformation Perspective
Authors: Joonkyu Kim, Yejin Kim, Jy-Yong Sohn
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Third, we support our theoretical findings through experiments on real image datasets, including Split-CIFAR100 and Image Net1K. |
| Researcher Affiliation | Academia | 1Department of Electrical & Electronic Engineering, Yonsei University, Seoul, South Korea 2Department of Statistics and Data Science, Yonsei University, Seoul, South Korea. |
| Pseudocode | No | The paper includes mathematical derivations and proofs but does not contain a clearly labeled pseudocode or algorithm block. |
| Open Source Code | No | The paper does not contain any explicit statements about the availability of open-source code, nor does it provide any links to code repositories. |
| Open Datasets | Yes | We test on Split-CIFAR100 dataset (Ramasesh et al., 2020) and a downsampled version of the original Image Net1K dataset (Chrabaszcz et al., 2017). |
| Dataset Splits | No | The paper mentions that for each dataset, the classes are partitioned into N = 50 categories, with each category containing 2 classes for Split-CIFAR100 and 5 classes for Image Net1K. It also mentions training a linear classifier for task t=1 on extracted features. However, it does not provide specific training/validation/test split percentages or absolute sample counts for the datasets themselves. |
| Hardware Specification | No | The paper describes the experimental setup, including datasets and model architecture, but does not specify any particular hardware components such as GPU or CPU models used for the experiments. |
| Software Dependencies | No | The paper mentions "OPTIMIZER ADAMW" in Table 1, but does not provide specific version numbers for this or any other key software libraries or frameworks used in the experiments. |
| Experiment Setup | Yes | Table 1. Continual learning hyper-parameter for Image Net32 training PARAMETER VALUE LEARNING RATE 0.001 BATCH SIZE 512 EPOCHS 500 WARM UP STEPS 200 WORKERS 4 OPTIMIZER ADAMW WEIGHT DECAY 5E-4 |