Consolidation using Multiple Task Learning with Context Inputsand Replay of CVAE Generated Pseudo-Examples
Bibliographic record
Abstract
A key component for a Lifelong Learning Agent is the integration or consolidation of new task knowledge with prior task knowledge. Consolidation requires a solution to several problems, most notably, the catastrophic forgetting problem where the development of representation for a new task reduces the accuracy of prior tasks. This paper extends our prior work on consolidation using multiple tasks learning (MTL) networks and a task rehearsal or replay approach. The goal is to maintain functional stability of the MTL network models for prior tasks, while providing representational plasticity to integrate new task knowledge into the same network. Our approach uses (1) a conditional variational autoencoder (CVAE) to generate accurate pseudo-examples (PEs) of prior tasks, (2) sweep-rehearsal requiring only a small number of PEs for each training iteration, (3) the appropriate weighing of PEs to ensure consolidation of new task knowledge with prior, and (4) a novel network architecture we call MTL with Context inputs (MTLc) which combines the best of standard MTL and context-sensitive MTL (csMTL) architectures. Sequential learning of twenty classification tasks using a combination of MNIST and Fashion-MNIST datasets shows that our CVAE based approach to generating accurate PEs is promising and that MTLc performs better than either MTL or csMTL with minimal loss of task accuracy over the sequence of tasks.
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 | 0.001 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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".