ONNLab is a framework and project for developing obfuscated neural networks, offering full control over training and inference, flexible architecture configuration, and seamless integration of custom components. Its core principle is systematic yet highly flexible design, enabling implementation of any architecture or training strategy.

Code for a Monte Carlo approximation of the prevalence of transitivity-failure in Bayesian confirmation. Other inference patterns are also covered.