▼ Step 1: Choose Data Source

Start with new experiments (LHS)

Upload existing experimental data

Upload Your Data File

Upload an Excel file (.xlsx, .xls) containing your experimental data. We'll automatically detect the columns and help you configure the optimization variables.

Select Excel File:

Data Preview (First 5 Rows):

Select Target/Objective Column:

▼ Step 2: Configure Optimization Variables

Optimization Goal

Select Goal: ? Choose whether you want to minimize or maximize your objective function. This affects how the optimizer interprets "better" results.

Advanced Optimizer Settings

Acquisition Function: ? Acquisition function determines how the optimizer selects next points to evaluate:

• Thompson Sampling: Bayesian approach using posterior sampling
• Expected Improvement: Balances exploration and exploitation
• Upper Confidence Bound: Optimistic approach with confidence intervals
• Probability of Improvement: Focuses on probability of finding better points
• Min/Max Predicted Value: Greedy approach using GP predictions

EI Jitter (xi): ? Jitter parameter for Expected Improvement. Higher values encourage more exploration by adding noise to the improvement calculation. Typical range: 0.01-0.1. This parameter adapts automatically based on optimization progress.

UCB Kappa (κ): ? Controls exploration vs exploitation trade-off in Upper Confidence Bound. Higher values favor exploration. Default 1.96 corresponds to 95% confidence interval. This parameter adapts automatically based on optimization progress.

PI Tau (τ, scaled space): ? Trade-off parameter for Probability of Improvement. Defines the minimum improvement threshold in scaled space. Smaller values are more greedy.

CV folds: ? Number of cross-validation folds for hyperparameter tuning of the Gaussian Process model. Higher values give more robust tuning but increase computation time.

Training Epochs: ? Maximum number of hyperparameter optimization evaluations for the GP model. Higher values may find better hyperparameters but require more computation time.

Early Stopping: ? Number of iterations without improvement before stopping GP hyperparameter optimization early. Prevents overfitting and reduces training time.

Thompson Samples: ? Number of posterior samples to draw for Thompson Sampling acquisition function. More samples provide better approximation but increase computation cost.

Multiple Recommendations Strategy: ? Strategy for generating multiple recommendations with Thompson Sampling:

• Independent: Each recommendation sampled independently (recommended)
• Top-K: Select top K points from single sampling
• Diverse: Encourage diversity among recommendations

Number of Recommendations per Step: ? Number of parameter suggestions to generate at each optimization step. Multiple recommendations allow parallel experimentation but may reduce individual quality.

▼ Step 3: Choose Candidate Point Strategy

The acquisition function needs candidate points to evaluate and recommend the next experiment. Choose your strategy:

🎲 Generate Random Candidates (Default)

Automatically generate random candidate points within your variable ranges for acquisition function evaluation.

📁 Upload Custom Candidates

Upload your own candidate points to restrict the search space to specific points of interest.

Number of Random Candidates: ? Number of random candidate points to generate for acquisition function optimization. Higher values improve optimization quality but increase computation time. Recommended: 1000-5000.

Upload a file containing candidate points. The file should contain columns matching your configured variables exactly.

Initial Experiments Setup

Please run experiments for the following initial parameter sets and record the objective value for each.

Upload Existing Experimental Data (Optional)

If you have existing data, upload an Excel file (.xlsx, .xls). Ensure column names in your file match the 'Feature Name' defined in Step 1.

Select Excel File:

Uploaded Data Preview (First 5 Rows):

Target/Objective Column Name:

Bayesian Optimization Step

Training Gaussian Process model, please wait...

This might take a few moments, especially with more data or higher hyperparameter evaluations.

Current Suggested Parameters:

N/A

Enter Observed Value for these parameters:

Selected Experiments and Results

Enter the observed values for your selected recommendations:

▼ Step 1: Choose Data Source

Upload Your Data File

Data Preview (First 5 Rows):

▼ Step 2: Configure Optimization Variables

Optimization Goal

Advanced Optimizer Settings

▼ Step 3: Choose Candidate Point Strategy

Initial Experiments Setup

Upload Existing Experimental Data (Optional)

Uploaded Data Preview (First 5 Rows):

Bayesian Optimization Step

Current Suggested Parameters:

Selected Experiments and Results

Objective Value History

GP Model Performance on Observed Data

Experiment Exploration in Feature Space

Optimization Status:

🎯 Convergence Analysis

⚙️ Adaptive Parameters

📈 Optimization Progress

Regret Evolution

Evaluation History