Class AdaptiveFitDetectorOptions
Configuration options for the Adaptive Fit Detector, which automatically selects the most appropriate method to detect overfitting and underfitting in machine learning models.
public class AdaptiveFitDetectorOptions- Inheritance
-
AdaptiveFitDetectorOptions
- Inherited Members
Remarks
The Adaptive Fit Detector combines multiple strategies to determine if a model is properly fitted to the data, overfitted (too complex, memorizing training data), or underfitted (too simple, missing patterns). It dynamically selects the most appropriate detection method based on the model's complexity and performance.
For Beginners: Think of this as a smart diagnostic tool that checks if your AI model is learning properly. Just like a doctor might use different tests depending on your symptoms, this detector chooses the right method to check if your model is learning too much detail from your data (overfitting) or not learning enough (underfitting). It automatically picks the best testing approach based on how complex your model is and how well it's performing.
Properties
ComplexityThreshold
Gets or sets the threshold that determines when to switch between different fit detection methods based on model complexity.
public double ComplexityThreshold { get; set; }Property Value
- double
The complexity threshold, defaulting to 1.0.
Remarks
Models with complexity scores above this threshold will be analyzed using methods better suited for complex models, while simpler models will use methods optimized for low-complexity scenarios.
For Beginners: This is like a dividing line that helps decide which testing method to use. If your model is more complex than this threshold (has more parameters or layers), the detector will use certain testing methods. If it's simpler, it will use different methods. Think of it like choosing different tools depending on whether you're working on a bicycle or a car - the complexity determines which tools are most appropriate.
HybridOptions
Gets or sets the configuration options for the Hybrid method of fit detection.
public HybridFitDetectorOptions HybridOptions { get; set; }Property Value
- HybridFitDetectorOptions
The hybrid options, initialized with default values.
Remarks
The hybrid approach combines multiple detection methods and metrics to provide a more robust assessment of model fit quality.
For Beginners: This is like getting a second opinion by combining multiple testing methods. Rather than relying on just one way to check if your model is learning properly, the hybrid approach uses several methods together to get a more complete picture. It's similar to how doctors might use both an X-ray and blood tests to make a more accurate diagnosis.
LearningCurveOptions
Gets or sets the configuration options for the Learning Curve method of fit detection.
public LearningCurveFitDetectorOptions LearningCurveOptions { get; set; }Property Value
- LearningCurveFitDetectorOptions
The learning curve options, initialized with default values.
Remarks
Learning curve analysis compares performance on training and validation data across different training set sizes to identify overfitting or underfitting.
For Beginners: This method tracks how well your model learns as you give it more training examples. If your model does great with the training data but poorly with new data, it's probably overfitting (like memorizing answers instead of understanding concepts). If it performs poorly on both training and new data, it's likely underfitting (like using too simple a formula to solve a complex problem).
PerformanceThreshold
Gets or sets the threshold that determines when to switch between different fit detection methods based on model performance.
public double PerformanceThreshold { get; set; }Property Value
- double
The performance threshold, defaulting to 0.8.
Remarks
Models with performance scores above this threshold will be analyzed using methods better suited for high-performing models, while lower-performing models will use methods that can better diagnose fundamental learning issues.
For Beginners: This is another dividing line, but based on how well your model is performing. If your model's accuracy (or other performance measure) is above this threshold (e.g., 80% correct), the detector will use certain testing methods. If it's performing worse, it will use different methods that are better at finding basic problems. It's like how a coach might focus on advanced techniques for a skilled athlete, but work on fundamentals with a beginner.
ResidualAnalysisOptions
Gets or sets the configuration options for the Residual Analysis method of fit detection.
public ResidualAnalysisFitDetectorOptions ResidualAnalysisOptions { get; set; }Property Value
- ResidualAnalysisFitDetectorOptions
The residual analysis options, initialized with default values.
Remarks
Residual analysis examines the differences between predicted and actual values to detect patterns that indicate overfitting or underfitting.
For Beginners: This method looks at the errors your model makes (the difference between what it predicts and the actual correct answers). If these errors show patterns instead of being random, it suggests your model isn't learning correctly. Think of it like checking if a student's test mistakes are random or if they consistently make the same type of error, which would indicate a misunderstanding.