Is there a way that, given a sample set of random values, using machine learning techniques one can be able to predict its probability distribution. What I mean is that, if I generate different sample sets, each drawn from different distributions, i.e. weibull, gamma, exp or lognormal, I can use these sets as training ones and then feeding some new data sets I would be able to find out their distribution as the best fit(if they fall into one of the above) ?
It would be very helpful if anyone could indicate anyway how to achieve this.
