Skip to main content

​
NimbusProbit — Bayesian Multinomial Probit Regression

Julia: NimbusProbit | Python: Not currently available
Mathematical Model: Bayesian Multinomial Probit Regression NimbusProbit is NimbusSDK.jl’s most flexible static classifier. Compared to Gaussian models (NimbusLDA, NimbusQDA), it can represent non-Gaussian decision boundaries while still producing calibrated posterior probabilities.
Availability
  • Julia SDK: ✅ NimbusProbit
  • Python SDK: ❌ Not currently available (use NimbusSoftmax for a non-Gaussian static model in Python)

​
Quick Start (Julia)

using NimbusSDK

# One-time setup
NimbusSDK.install_core("nbci_live_your_key")

# Train
model = train_model(
    NimbusProbit,
    train_data;
    iterations = 50
)

# Batch inference
results = predict_batch(model, test_data)
println("Accuracy: $(sum(results.predictions .== labels) / length(labels))")

​
When to Use NimbusProbit

  • Complex multi-class tasks where NimbusLDA / NimbusQDA plateau
  • Non-Gaussian boundaries (e.g., overlapping clusters not well modeled by class-conditional Gaussians)
  • You still need uncertainty quantification and stable probabilities

​
When Not to Use It

  • If latency must be minimized: start with NimbusLDA, then NimbusQDA
  • If the task is non-stationary / drifting: use NimbusSTS (Python-only)

​
Next Steps

  • Julia SDK reference: /julia-sdk/api-reference
  • Static Gaussian models: /models/rxlda, /models/rxgmm
  • Python non-Gaussian static model: /models/rxpolya