AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Decision-Focused Learning via Tangent-Space Projection of Prediction Error

Junhyeong Lee, Sangjin Jin, Yongjae Lee

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Active learning and subset selection · 0 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 PEAR tangent-space regret gradient instead of two-stage MSE traininglinear predictor · Shortest Path (degree 8)normalized regret (%): 15.428 -> 4.246Improved
use PEAR instead of SPO+linear predictor · Shortest Path (degree 4)normalized regret (%): 0.761 vs 0.774No clear effect
use PEAR instead of SPO+ for training timelinear predictor · Shortest Path (degree 4)training time (s): 41.8 -> 23.5Improved
use PEAR instead of SPO+linear predictor · Knapsack (degree 8)normalized regret (%): 0.763 -> 0.437Improved
use PEAR instead of LAVAlinear predictor · Shortest Path (degree 8)normalized regret (%): 6.189 -> 4.246Improved
use PEAR instead of LAVAlinear predictor · Knapsack (degree 8)normalized regret (%): 1.351 -> 0.437Improved
use PEAR instead of LAVAlinear predictor · Knapsack (degree 8)training time (s): 182.8 -> 271.7Hurt
use PEAR instead of SPO+ under multiplicative noise 0.5linear predictor · Shortest Path (degree 8, noise 0.5)normalized regret (%): 10.49 -> 10.38No clear effect
use PEAR instead of SPO+ under multiplicative noise 0.1linear predictor · Knapsack (degree 8, noise 0.1)normalized regret (%): 0.81 -> 0.49Improved
use PEAR instead of QPTH layerDLinear · Real-world QP (S&P 100 MVO)normalized regret (%): 87.85 -> 85.38Improved
use PEAR instead of CvxpyLayersDLinear · Real-world QP (S&P 100 MVO)training time (s): 321.9 -> 122.3Improved
use PEAR instead of MSEDLinear · Real-world QP (S&P 100 MVO)Sharpe ratio: 0.92 -> 1.44Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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