AutoMLSearchStrategy
Enums · AiDotNet.Enums
Defines the search strategy used to explore AutoML candidate configurations.
For Beginners
This is how AutoML decides what to try next:
RandomSearchtries random settings (simple and surprisingly strong).BayesianOptimizationtries to learn which settings work best and focus on them.Evolutionaryevolves good settings over time (useful for discrete/conditional knobs).MultiFidelityuses short runs first and only gives more budget to promising trials.MapEliteskeeps strong but behaviorally diverse model configurations.
How It Works
AutoML can use different strategies to decide which candidate model configurations to try next. The best choice depends on budget, search-space shape (continuous vs categorical), and how expensive each trial is.
Fields
| Field | Summary |
|---|---|
AutoMLSearchStrategy BayesianOptimization |
Bayesian optimization (typically Gaussian-process or TPE style). |
AutoMLSearchStrategy DARTS |
DARTS (Differentiable Architecture Search) - gradient-based NAS. |
AutoMLSearchStrategy Evolutionary |
Evolutionary / genetic search. |
AutoMLSearchStrategy GDAS |
GDAS (Gumbel-softmax DARTS) - improved differentiable NAS. |
AutoMLSearchStrategy MapElites |
MAP-Elites quality-diversity search over model family and configuration complexity. |
AutoMLSearchStrategy MultiFidelity |
Multi-fidelity search (e.g., HyperBand/ASHA-style scheduling). |
AutoMLSearchStrategy NeuralArchitectureSearch |
Neural Architecture Search with automatic algorithm selection. |
AutoMLSearchStrategy OnceForAll |
Once-for-All (OFA) Networks - train once, specialize anywhere. |
AutoMLSearchStrategy RandomSearch |
Random search baseline. |
int value__ |