An implementation of Bayesian Flow Networks
Algorithm 5 of the paper has been implemented [Continuous-Time Loss L∞(x) for Discretised Data]. I am looking forward to add all, but I'd be happy if you can contribute.
The model has been tested with toy data with sinousoidal shape . Here is the generated sample versus a sample from dataset.
