Calculation Results
Your search-space estimate appears here.
Values per Hyperparameter
Runtime Comparison
Search Strategy Comparison
Combination Preview
Strategy Comparison
| Strategy | Combinations | Model Fits | Sequential Time | Estimated Cost |
|---|
Search-Space Inputs
Add values, ranges, logarithmic scales, and conditional dependencies.
Hyperparameters
Validation and Repetition
Runtime and Parallel Resources
Memory, Storage, and Cost
Editable Warning Thresholds
Calculation History
Recent calculations are stored in this browser.
| Date | Parameters | Combinations | Fits | Parallel Time | Cost |
|---|
Formula Used
Total combinations = V₁ × V₂ × ... × Vₙ. Conditional parameters contribute values only when their rules match.
Total model fits = selected combinations × folds × repeats × models. Failed-run retries add expected replacement fits.
Parallel time = sequential time ÷ workers ÷ efficiency. Actual scheduling and hardware overhead may change runtime.
Estimated cost uses compute hours and per-fit charges. Storage and memory figures support planning.
How to Use
- Select a preset or create custom hyperparameters.
- Enter lists, numeric ranges, or logarithmic ranges.
- Add conditions when a parameter depends on another.
- Choose a search strategy and validation method.
- Enter runtime, workers, costs, and resource assumptions.
- Calculate, review warnings, and export configurations.
Example Data
| Hyperparameter | Candidate values | Count |
|---|---|---|
| Learning rate | 0.001, 0.01, 0.1 | 3 |
| Batch size | 16, 32, 64 | 3 |
| Optimiser | Adam, SGD | 2 |
| Total grid combinations | 18 | |
Frequently Asked Questions
What is a hyperparameter combination?
It is one complete set of chosen tuning values. Each set defines one model configuration.
Why do combinations grow quickly?
Every additional value multiplies the search space. Several small lists can create thousands of configurations.
What is the difference between grid and random search?
Grid search evaluates every selected combination. Random search evaluates a limited sample.
How does cross-validation affect model fits?
Each combination is trained once per fold. More folds increase reliability and computation.
What are conditional hyperparameters?
They become active only after another parameter matches a rule. Kernel degree is a common example.
Does parallel processing reduce total compute?
It mainly reduces elapsed time. Total compute hours may remain similar or increase slightly.
How should I choose random-search iterations?
Use budget, runtime, and parameter importance. Begin small, then expand promising regions.
Why can actual runtime differ?
Data loading, hardware, caching, failures, and model complexity change training time.
Can I export generated configurations?
Yes. Export CSV, JSON, PDF, or framework-oriented configuration files.