What data is needed for machine learning based on RNG prediction?

Question

Neural networks

Recently, the following problem has arisen. There is a random number generator (naturally pseudo-random). Numbers are distributed in the range from 1 -1000 inclusive. Next, a dataset was collected, in which about 45000 + -2000, these numbers were randomly generated. Is it possible to train a neural network in such a way that it predicts the next number? Or is more data needed, not the number of numbers themselves, but the input variables on which the output number depends?

Answer 1

Neural network - hardly. Her main profile is working with smooth continuous quantities. And if we are talking about prediction, then here the term is rather close - approximation. Or extrapolation.
And if your RNG is crypto-resistant, then consider this a hopeless case. It was created by a specialist so that no one would ever guess about the forecasts.
RNG is something discrete. Here, combinatorics and GA are more suitable. For example, just check if your dataset belongs to some class of values. By repetition. Distribution form. If it's not linear.

Answer 2

In your case, the task comes down to predicting the quantity. First, try to solve it using "classic" machine learning methods (for example, linear regression). Link to blitz check . And there the need for ANN will disappear.)