RiPPLE: Cross-Space Performance Prediction from Early Training for Neural Architecture Search

arXiv:2609.12418v1 Announce Type: cross Abstract: Neural architecture search (NAS) evaluates candidate networks, but fully training enough architectures to rank an entire space is expensive. Zero-cost proxies score architectures at initialization, yet their ranking quality varies across search spaces. Learned predictors reduce evaluation cost but typically require fully trained labels or partial-training features for individual candidates. We introduce extbf{RiPPLE}, nderline{extbf{R}}anking vnderline{extbf{i}}a nderline{extbf{P}}refix-nderline{extbf{P}}ropagated nderline{extbf{L}}abel nderline{extbf{E}}xtrapolation, which treats partial training as a source of labels for a small coverage set of anchors. RiPPLE trains these anchors to an early prefix, extrapolates their learning curves to surrogate labels, and propagates the labels over label-free architecture features. The early-training signal remains a label on the anchors rather than a per-candidate feature. Feature, readout, and encoding rules are selected without held-out accuracy and reused across search spaces. We evaluate the method on twelve benchmark cells from four search-space families and on the larger DARTS space. The results examine ranking quality, label efficiency, architecture selection, and the roles of readout, coverage, and propagation. RiPPLE provides a whole-space ranking from a fractional anchor-training budget, with comparisons interpreted under their respective evaluation and cost protocols.

Related

• CoRA-NAS: Coarse Ranking and Anchor-Residual Refinement for Neural Architecture Search
• Bi-EZP: LLM-Guided Bilevel Program Evolution for Ensemble Zero-Cost Proxy Discovery
• ONNX-Net: Towards Universal Representations and Instant Performance Prediction for Neural Architectures

Source: arXiv cs.CV | 2026-09-14