Continuous Histogram Computing on Streaming Data Approach

In federative distributed learning, when combining trained models into one, information about the distribution of training samples is used. However, in some cases (for example, when training with reinforcement), it is impossible to obtain information about the distribution using standard methods due to the large volume of data or their generation during training. This paper describes the method of computing a histogram for the case of a random value with a previously unknown min and max values in one pass through the streaming data.