AI Paper Atlas

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

Learning to Optimize on SPD Manifolds

Zhi Gao, Yuwei Wu, Yunde Jia, Mehrtash Harandi

CVPR 2020 · Learning theory and algorithms › Prediction and decision losses · approach: Determinantal point process inference · 24 citations (OpenAlex)

Check 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 the learned SPD optimizer rather than the strongest reported hand-designed optimizerSPD optimization · Kylbergclassification accuracy: 84.37 vs 83.19Improved
Use the learned SPD optimizer rather than the strongest reported hand-designed optimizerSPD optimization · CUBclassification accuracy: 70.97 vs 70.44Improved
Replace conventional LSTM with the symmetric-structure-preserving mLSTM (ablation)Learned SPD optimizer · Kylberg clusteringlog loss: 2.47 and 1.39 without mLSTM; 0.16 and -0.77 for oursImproved
Add experience replay and an experience pool (ablation)Learned SPD optimizer · Kylberg clusteringconvergence rateImproved
Add a skip-connection to fit the residual search direction (ablation)Learned SPD optimizer · Kylberg clusteringconvergenceImproved
Use the learned SPD optimizer instead of RSGDSPD optimization · MNISToptimization time: 76.757 vs 74.980Hurt
Use the learned SPD optimizer instead of RSGDSPD optimization · Kylbergoptimization time: 378.996 vs 367.995Hurt
Use the learned SPD optimizer instead of RSGDSPD optimization · CUBoptimization time: 154.370 vs 150.262Hurt

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