Skip to main content

Parameter Space Analysis

The Parameter Space view in InForm provides a powerful way to visualize and understand the entire design space, helping you identify patterns, optimal regions, and parameter relationships.

Understanding Parameter Space​

What is Parameter Space?​

Parameter space is a multi-dimensional representation where:

  • Each dimension represents one design parameter
  • Each point represents a unique design configuration
  • Distance between points indicates similarity of designs
  • Clusters show regions of similar performance

Visual Representation​

InForm uses several visualization techniques:

  • Scatter plots: 2D/3D projections of the parameter space
  • Parallel coordinates: All parameters shown simultaneously
  • Heat maps: Performance overlaid on parameter combinations
  • Density plots: Show concentration of design points

Accessing the View​

  1. Open your project in InForm
  2. Click on the "Parameter Space" tab in the main navigation
  3. Wait for the data to load and visualize
  4. Use the controls to adjust the view

Interface Elements​

Visualization Controls​

  • Axis selection: Choose which parameters to display
  • Color mapping: Map performance metrics to colors
  • Point size: Adjust based on another metric
  • Filter controls: Focus on specific parameter ranges

Interaction Tools​

  • Selection: Click and drag to select design points
  • Zoom: Scroll to zoom into regions of interest
  • Pan: Drag to move around the space
  • Brush: Highlight ranges of interest

Visualization Techniques​

Scatter Plot View​

2D Scatter Plots​

  • X and Y axes: Select two key parameters
  • Color coding: Show performance with color intensity
  • Point clustering: Identify regions of similar designs
  • Trend lines: Show correlations between parameters

3D Scatter Plots​

  • Additional dimension: Add a third parameter
  • Interactive rotation: Explore from different angles
  • Depth perception: Understand 3D relationships
  • Performance surfaces: Visualize how performance varies

Parallel Coordinates​

Understanding the Display​

  • Vertical lines: Each represents one parameter
  • Horizontal position: Parameter value along each line
  • Connected lines: Show individual design configurations
  • Line color: Indicates performance level

Interactive Features​

  • Brushing: Select ranges on any parameter
  • Filtering: Hide designs outside selected ranges
  • Reordering: Drag parameter lines to change order
  • Inversion: Flip parameter scales if needed

Heat Maps and Density Plots​

Performance Heat Maps​

  • Grid representation: Divide parameter space into cells
  • Color intensity: Shows average performance in each cell
  • Hot spots: Identify high-performing regions
  • Cold zones: Areas to avoid

Density Visualization​

  • Point concentration: Shows where most designs cluster
  • Sparse regions: Areas with few explored designs
  • Coverage analysis: Understand exploration completeness

Analysis Techniques​

Pattern Recognition​

  1. Look for diagonal patterns in scatter plots (correlations)
  2. Find curved relationships (non-linear effects)
  3. Spot clusters of similar designs
  4. Identify outliers that behave differently

Performance Landscapes​

  • Peaks: Regions of optimal performance
  • Valleys: Poor performing areas
  • Ridges: Lines of good performance
  • Plateaus: Areas with similar performance

Statistical Analysis​

Correlation Analysis​

  • Strong correlations: Parameters that move together
  • Inverse correlations: Parameters that oppose each other
  • Independence: Parameters with no relationship
  • Non-linear relationships: Complex parameter interactions

Distribution Analysis​

  • Parameter ranges: Understand the explored space
  • Performance distribution: Spread of objective values
  • Constraint satisfaction: Feasible vs. infeasible regions

Advanced Features​

Multi-objective Analysis​

Pareto Frontier​

  • Definition: Set of non-dominated solutions
  • Visualization: Often shown as a curve or surface
  • Trade-off analysis: Understand competing objectives
  • Selection: Choose preferred trade-offs

Objective Space View​

  • Switch axes: From parameters to objectives
  • Performance comparison: Direct comparison of goals
  • Constraint boundaries: Show feasible regions
  • Optimal sets: Identify best performing clusters

Filtering and Selection​

Dynamic Filtering​

  1. Set parameter ranges using sliders or input fields
  2. Apply performance thresholds to focus on good designs
  3. Combine multiple filters for complex selections
  4. Save filter sets for later use

Selection Tools​

  • Lasso selection: Draw around interesting regions
  • Box selection: Select rectangular areas
  • Multi-selection: Add to existing selections
  • Invert selection: Select everything except current

Data Export and Analysis​

Export Options​

  • Selected points: Export specific design configurations
  • Full dataset: Export all explored designs
  • Performance data: Include objective values and constraints
  • Parameter metadata: Include descriptions and units

External Analysis​

  • CSV format: Import into Excel or other tools
  • Statistical software: Use R, Python, or MATLAB
  • Custom analysis: Build specialized analysis tools
  • Reporting: Create summary reports and presentations

Best Practices​

Effective Exploration​

Systematic Sampling​

  1. Start with design of experiments (DoE) approaches
  2. Use space-filling designs for broad coverage
  3. Adaptive sampling: Focus on interesting regions
  4. Validate with additional points

Iterative Refinement​

  • Coarse exploration first: Understand the big picture
  • Zoom into promising regions: Add detail where needed
  • Balance breadth and depth: Don't focus too narrowly
  • Document findings: Keep track of insights

Visualization Best Practices​

Choosing Views​

  • Start with parallel coordinates for overview
  • Use scatter plots for detailed relationships
  • Apply heat maps for performance landscapes
  • Switch between views to gain different insights

Color and Scaling​

  • Use intuitive color maps: Red for bad, green for good
  • Normalize scales: Ensure fair comparison between parameters
  • Highlight extremes: Make outliers visible
  • Maintain consistency: Use same colors across views

Common Analysis Patterns​

Design Space Characterization​

Complete Exploration​

  1. Uniform sampling: Regular grid of parameter values
  2. Random sampling: Monte Carlo exploration
  3. Latin hypercube: Efficient space-filling design
  4. Halton sequences: Low-discrepancy sampling

Targeted Exploration​

  1. Optimization-driven: Follow gradients toward optima
  2. Constraint-focused: Explore boundary regions
  3. Sensitivity-based: Focus on influential parameters
  4. User-guided: Interactive exploration based on insights

Performance Analysis​

Single Objective​

  • Global optimization: Find the single best solution
  • Robustness analysis: Find solutions that perform well across variations
  • Constraint analysis: Understand feasible regions
  • Sensitivity study: Identify critical parameters

Multi-objective​

  • Pareto analysis: Find the trade-off frontier
  • Preference exploration: Understand stakeholder priorities
  • Compromise solutions: Find balanced designs
  • Scenario analysis: Understand performance under different conditions

Troubleshooting​

Visualization Issues​

  • Too many points: Use sampling or filtering to reduce density
  • Overlapping data: Adjust transparency or point size
  • Scale problems: Normalize or transform parameter values
  • Missing data: Check for incomplete evaluations

Analysis Difficulties​

  • No clear patterns: Try different parameter combinations or transformations
  • Conflicting objectives: Accept that trade-offs are necessary
  • Constraint violations: Understand feasible regions better
  • Performance plateaus: Look for small improvements or different objectives

Next Steps​