atoum84

Hello, I have a large ode problem and I am also interrested by a gpu feature.

I'm trying to implement the cross entropy objective (multi-dimension): -1/n * sum( y_expected * log(y_prediction) + (1-y_expected) * log(1-y_prediction) ) https://en.wikipedia.org/wiki/Cross_entropy

Yes, my first idea was to adapt the code of LogLikelihood. However, in the future, I will need to implement a custom objective that's why I choose MinimizeModel which seems...

The source code for the cross entropy objective: ``````python class CrossEntropy(GradientObjective): """ Error function to be minimized by a minimizer in order to *maximize* the cross-entropy. """ @keywordonly(flatten_components=True) def __call__(self,...

You can do a simple classifier. For exemple: + You have a 1D model: y = np.clip(x < k, eps, 1.0-eps) (k is the parameter to fit, eps=1E-6) + A...