Forget Less, Learn More: Contrastive-Based Federated Class Incremental Learning with a Low-Dimensional Projection Layer
Bibliographic record
Abstract
Federated Class Incremental Learning (FCIL) extends Federated Learning (FL) to dynamic environments where clients continually encounter new classes over time, but past data becomes inaccessible. This leads to catastrophic forgetting of previous classes. Therefore, maintaining a balance between learning new classes (plasticity) and retaining past knowledge (stability) is crucial. To address these challenges, we propose a contrastive-based FCIL framework with a low-dimensional projection layer to enhance both stability and plasticity. A low-dimensional projection layer is introduced to improve the generalization capability of the feature extractor and stability-plasticity trade-off. In this regard, we employ supervised contrastive learning in the projection layer during the initial task to strengthen the feature extractor for better adaptation to new classes. Additionally, we train a data-free generator on the server and distribute it to clients to replay synthetic samples from past tasks. To balance stability and plasticity, we design a novel loss function that integrates cross-entropy loss, feature-level knowledge distillation loss, contrastive loss in the projection layer, and classification head refinement. Extensive experiments demonstrate that our framework outperforms state-of-the-art FCIL baselines, achieving higher accuracy and lower forgetting. The code is available at https://github.com/EnsiehKhazaei/FCIL-LD.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.004 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".