hi, i want to know whether multi-layer Bayesian neural network and MLP(Multi-Layer perceptron), with Bayesian regularisation same or different??

hi, i want to know whether multi-layer Bayesian neural network and MLP(Multi-Layer perceptron), with Bayesian regularisation same or different??

Answer 1

[Greg Heath]

They are different.

 However, I will let you run both to see if the same inputs yield the same outputs.

 It will help if you

1. Start each at the same state of the RNG.

2. Remove line ending semicolons to view relevant results.

Good Luck!

Thank you for formally accepting my answer.

Greg

PS: The help and doc online documentation for trainbr, the obsolete msereg and mse show that the regularization factor in trainbr varies during training to optimize the mse(weights)/mse(error) ratio. Otherwise, the regularization factor of msereg seems to be fixed with a default value of 0.05 that can be changed and the regularization factor of mse is set by the designer.