AI Paper Atlas

Research map / Optimization / Optimizers

Statistical Learning and Inverse Problems: A Stochastic Gradient Approach

Yuri Fonseca, Yuri Saporito

NeurIPS 2022 · Optimization › Optimizers · approach: Stochastic gradient MCMC · 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 20 spline degrees of freedom instead of 10 (ablation)ML-SGD with spline base learner · Bitcoin darknet classificationavg accuracy: 0.73 -> 0.79Improved
use 20 spline knots instead of 10 (ablation)FPLR · Bitcoin darknet classificationavg accuracy: 0.74 -> 0.80Improved
replace spline base learner with regression tree (depth 20) (ablation)ML-SGD · Bitcoin darknet classificationavg accuracy: 0.79 -> 0.78No clear effect
plain SGD-SIP vs ML-SGD with spline k=20 (ablation)SGD-SIP · Bitcoin darknet classificationavg accuracy: 0.79 -> 0.80No clear effect
SGD-SIP vs tailored FPLR (k = 20)SGD-SIP · Bitcoin darknet classificationavg accuracy: 0.80 -> 0.80No clear effect
smoothing-spline base learners in place of raw stochastic gradientsML-SGD · Synthetic FLR and deconvolutionMSE / smoothnessMixed
compare SGD-SIP/ML-SGD with Landweber iterationSGD-SIP · Synthetic FLR (sine)MSENo clear effect
compare ML-SGD with tailored PFLR on smooth sine targetML-SGD · Synthetic FLR (sine)MSEHurt
use a non-smooth step target functionPFLR advantage over SGD methods · Synthetic FLR (step)MSEHurt
use all samples at every iteration instead of oneSGD-SIP / ML-SGD · Synthetic FLR (step)MSEImproved

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