How ServAdvisor Works — Artificial Intelligence

AI and Stochastic optimization part 1

In fact, many phenomena observed in the physical universe are actually best modeled with nonlinear transformations. We use this in ServAdvisor process modeling for transformations between system inputs and the target output in machine learning and AI solutions.

For AI model training and the optimization of stochastic parameters of models, we develop special adaptive Generic Algorithm (GA) involving the idea of randomness when performing a search. However, it must be clearly understood that GAs are not simply random search algorithms. They utilize knowledge from previous generations of strings in order to construct a new generation that will approach the optimal solution.

Summarizing, the following essential features of GAs can be listed:

ØGeneric algorithms manipulate structures which represent the parameters, not the actual values of the parameters themselves

ØGeneric algorithms use a population of points to perform a search, not just a single point on the parameter space

ØGeneric algorithms use only the current measure of ‘’’goodness’’ to guide themselves to the optimal solution

ØGeneric algorithms are probabilistic in nature, not deterministic

ØGeneric algorithms are inherently parallel, dealing with a large number of points (strings) simultaneously

Apparently, GAs transfer the biological mechanisms of reproduction, crossover, and mutation to algorithms.

Moreover, efficient Approximate Stochastic Maximum Likelihood Estimates (AMLEs) are used, where they are known to have asymptotically optimum properties. Furthermore, the elements of the Cramer-Rao bound (CRB) matrix will be considered as a lower bound of the error covariance matrix of the ML-estimates. This establishes relations between model parameters and efficiency of the ML-estimates.

#cryptonews #cryptocurrency #blockchain #ICO #Crypto #TokenSale #earlybird #bitcoin #cryptokitties #altcoin #ServAdvisor #SRV