AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Transfer Learning with Affine Model Transformation

Shunya Minami, Kenji Fukumizu, Yoshihiro Hayashi, Ryo Yoshida

NeurIPS 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Statistical transfer learning · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use affine model transfer on top of feature extraction with GPT-3 embeddingsfeature-extraction TL · SciRepEval review scoreRMSE: 1.3284 -> 1.3234Improved
use affine model transfer instead of feature extraction with SciBERT embeddingsfeature-extraction TL · SciRepEval review scoreRMSE: 1.2856 -> 1.2797Improved
use affine model transfer instead of feature extraction with BERT embeddingsfeature-extraction TL · SciRepEval review scoreRMSE: 1.3086 -> 1.3069Improved
use affine model transfer instead of feature extraction with GPT-3 embeddingsfeature-extraction TL · SciRepEval acceptanceaccuracy: 0.6279 -> 0.6386Improved
use affine model transfer instead of feature extraction with SciBERT embeddingsfeature-extraction TL · SciRepEval acceptanceaccuracy: 0.6520 -> 0.6507No clear effect
add source features via AffineTL-full vs Direct on a weakly related taskDirect kernel ridge · SARCOS Torque 1, n=5RMSE: 21.3 -> 21.2No clear effect
add source features via AffineTL-full vs Direct on a related taskDirect kernel ridge · SARCOS Torque 7, n=5RMSE: 2.66 -> 2.23Improved
use AffineTL-full instead of Direct at n=50Direct kernel ridge · SARCOS Torque 7, n=50RMSE: 1.05 -> 0.870Improved
use Only source features instead of DirectDirect kernel ridge · SARCOS Torque 1, n=5RMSE: 21.3 -> 24.0Hurt
use HTL-offset instead of Direct (negative transfer)Direct kernel ridge · SARCOS Torque 1, n=50RMSE: 10.8 -> 15.1Hurt
use Fine-tune instead of AffineTL-full when source-target relation is strongAffineTL-full · SARCOS Torque 7, n=50RMSE: 0.870 -> 0.790Improved
use PAC-Net instead of Direct on a weakly related task, n=50Direct kernel ridge · SARCOS Torque 1, n=50RMSE: 10.8 -> 17.1Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy