Understanding transport in magnetically confined plasmas is critical for developing predictive models for future devices such as ITER. Thanks to recent progress in simulation and theory, along with enhanced computational power and better diagnostic systems, direct and quantitative comparisons between experimental results and models is possible. However, validating transport models using additional constraints and accounting for experimental uncertainties still remains a formidable task. In this work, a new optimization framework is developed to address the issue of constrained validation of transport models. The Validation via Iterative Training of Active Learning Surrogates (VITALS) framework exploits surrogate-based strategies using Gaussian processes and sequential parameter updates to achieve the combination of plasma parameters that matches experimental transport measurements within diagnostic error bars. VITALS is successfully implemented to study L-mode plasmas in the Alcator C-Mod tokamak, and for the first time, additional measurable quantities, such as incremental diffusivity and fluctuation levels, are used during the validation process of the quasi-linear transport models TGLF-SAT1 and TGLF-SAT0. First results indicate that these machine-learning algorithms are very suitable and adaptable as a self-consistent, fast, and comprehensive validation methodology for plasma transport codes.