Adaptive RBF-KAN: A Comparative Evaluation of Dynamic Shape Parameters in Kolmogorov-Arnold Networks
2026-05-22T07:24:06Z•c54660ae9d0705f4d2f976c9d52474748b9c16a22a4f0a1138eb831641248209
HSICadvertisingbatch-scalingbehavioral-divergencecausal-inferencecovariate-selectiondeep-learningdegree-heterogeneityeigengapgraph-embeddingkernel-methodslaw-of-the-iterated-logarithmleave-one-out-cross-validationmachine-learningmartingale-test','mHSIC','mdHSIC'matérn-kernelmean-field-dynamicsoff-policy-evaluationpropagation-of-chaosradial-basis-functionsreinforcement-learningspectral-embeddingstatistical-testingsupport-aware-decisionswendland-kernel
What happened
Collection of ten new arXiv preprints (2026-05-22) across machine learning and statistics. Key contributions include: Adaptive RBF-KAN — extends Kolmogorov-Arnold Networks with data-driven LOOCV initialization and additional radial kernels (Matérn, Wendland) for adaptive shape parameters; a local covariate-selection method for unbiased average causal-effect estimation that removes pretreatment and causal sufficiency assumptions; Adaptive Batch Scaling (ABS) for on-policy RL using a Behavioral Divergence metric to scale batches and reconcile early plasticity with late-stage convergence; a "sort
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_stat_ml
- Record identifier
- c54660ae9d0705f4d2f976c9d52474748b9c16a22a4f0a1138eb831641248209
- Enrichment time
- 2026-05-22T07:24:06Z
- 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.