Federated Learning for Sentiment Analysis in Presence of Non-IID Data: Sensitivity of Deep Learning Models
Bibliographic record
Abstract
In sentiment analysis, data are commonly distributed across many devices, and traditional machine learning requires transferring these data to a central location exposing data to security and privacy risks. Federated Learning (FL) avoids this transfer by training a model without requiring the clients/devices to share their local data; however, FL performance drops when data are not Independent and Identically Distributed (non-IID), such as when label distribution or data size vary across clients. Although techniques for non-IID data have been proposed primarily in the image domain, the sensitivity of various deep learning models to non-IID data needs to be examined. Consequently, this paper investigates the sensitivity of three dominant techniques in sentiment analysis, feed-forward neural networks, LSTMs, and transformers to common types of non-IID data, specifically data size and label imbalances. The scenarios were designed with increasing degrees of imbalance in terms of data size and label distribution to investigate gradual changes. The results revealed that label imbalance has a higher impact on accuracy than data size imbalance irrelevant of the algorithm. Overall, the transformer achieved the highest accuracy, and, while all models experienced a drop in accuracy with the increased label imbalance, this drop was smaller for the transformer, making it well suited for non-IID data.
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 | 0.006 | 0.025 |
| 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 | 0.002 | 0.004 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".