Counterfactual Explanations for Deep Two-Sample Testing
2026-06-04T07:24:02Z•aa31048720c624be8595eb15f92acb04eb864f0fa72bc0b55c06af33b637a72d
Expected-ShortfallMRI-analysisValue-at-Riskalternating-power-iterationautoencodersbayesian-reinforcement-learningcounterfactual-explanationscoupled-gradient-descentdeep-learningdiffusion-autoencoderfalse-discovery-ratefeature-selectionfinancial-riskflatness-generalizationforecast-reconciliationhomogeneous-networksknockoffsmachine-learningmaximum-mean-discrepancyoptimization-theorypseudospectral-analysis","anomaly-detection","time-series-anomalstatistical-testingstochastic-shortest-pathtensor-PCAtime-series-forecasting
What happened
This document is an arXiv stat_ml feed (multiple new papers) summarizing recent methods and theoretical results across machine learning and statistics. Key contributions include: a counterfactual-explanation framework for deep two-sample testing that uses a diffusion autoencoder + pretrained deep test and optimizes MMD in representation space to produce plausible sample edits (evaluated on synthetic 2D shapes and MRI cohorts); a finite-iteration local theory and warm-start principle for alternating power iteration in spiked tensor PCA; REGAIN, a reconciliation-gain method to learn and select
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_stat_ml
- Record identifier
- aa31048720c624be8595eb15f92acb04eb864f0fa72bc0b55c06af33b637a72d
- Enrichment time
- 2026-06-04T07:24:02Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.