Artificial Intelligence

2608 Submissions

[2] ai.viXra.org:2608.0095 [pdf] submitted on 2026-08-30 22:04:21

ZAPS: Zero-Cost Active Proxy Search for Neural Architecture Search

Authors: Hassan Touayouch, Rabie Najem, Mohammed Benjelloun
Comments: 9 Pages.

Neural Architecture Search (NAS) automates the design of neural networks but faces the prohibitive cost of evaluating eachcandidate architecture through full training. Zero-cost proxies estimate architecture quality in seconds without training, yetindividual proxies remain noisy, and no existing method jointly exploits proxy signals and the topological structure of architectureswithin an active-learning framework. We propose ZAPS (Zero-cost Active Proxy Search), a four-component pipeline that addressesthese limitations. ZAPS first selects a compact, non-redundant subset of proxies via PROXYFIT, a greedy anti-redundancy algorithm; initializes the search with a hybrid K-means strategy balancing exploitation and exploration; dynamically refines proxy selection using bootstrapped MRMR; and predicts architecture quality with an XGBoost ensemble fed by proxy ranks and one-hot topological encodings, guided by an Upper Confidence Bound (UCB) acquisition function. On NAS-Bench-201 with a budget ofB =200evaluations, ZAPS identifies 53.8% of the top-100 architectures on CIFAR-10 and 66.8% on CIFAR-100 (P@100), while maintaining a Regret of 0.04% on CIFAR-10. ZAPS outperforms comparable-budget methods — REA, BANANAS, and TPE — on both metrics evaluated on CIFAR-10 and CIFAR-100 of NAS-Bench-201.
Category: Artificial Intelligence

[1] ai.viXra.org:2608.0094 [pdf] submitted on 2026-08-30 18:48:10

Machine Learning: Foundations, Algorithms, Applications, Evaluation, and Future Directions

Authors: Samuelson G
Comments: 16 Pages. Just out of curiosity

Machine learning has rapidly advanced from early theoretical concepts to a practical technology that drives many modern applications (computer vision, natural language processing, etc.). We present a detailed survey covering ML foundations, algorithmic methodologies, and empirical evaluation. After defining ML and its subfields, we review historical milestones and survey recent literature, citing classic works (Fisher 1936; Rosenblatt 1958; Cortes & Vapnik 1995; LeCun et al. 2015) and modern treatments. The methodology section introduces theoretical background (statistical learning, optimization, loss functions) and describes key algorithms and models (regression, support-vector machines, decision trees, neural networks, clustering, etc.), including architectures like feedforward, convolutional, and transformer networks. We propose a generic ML workflow (Figure 4) using flowcharts. Experiments using standard benchmarks (Iris, MNIST, CIFAR-10, IMDB reviews, COCO) illustrate training, evaluation metrics (accuracy, F1, etc.), and dataset characteristics (Table 3). Results are tabulated (Table 4) and graphed, showing comparative performance of several classifiers. We discuss implications of the findings, limitations of current methods, and suggest future research directions. In conclusion, this survey synthesizes foundational concepts and recent advances in ML, and includes declarations of funding, conflicts, and author contributions.
Category: Artificial Intelligence