ArtificialEvolvingWorld / Entendre

A library of neural network implementations
0 stars 0 forks source link

Random.hh #4

Closed Lunderberg closed 7 years ago

Lunderberg commented 8 years ago

Rather than having each subtype be a different distribution, I think that all distributions should be available to the user. That is, something like the following base class.

class RNG {
public:
    virtual double uniform(double low, double high) = 0;
    virtual double gauss(double mean, double sigma) = 0;
};

That way, each class that uses RNG can use whichever distributions it needs, but we can still pass in a dummy RNG for testing purposes.

Lunderberg commented 7 years ago

Done as of fa4cc49