Incremental Learning EDA on One-Max

Incremental Learning implementation of Estimation of Distribution Algorithm (EDA) on one-max problem

What are EDAs?

Estimation of distribution algorithms (EDAs), sometimes called probabilistic model-building genetic algorithms (PMBGAs), are stochastic optimization methods that guide the search for the optimum by building and sampling explicit probabilistic models of promising candidate solutions. Optimization is viewed as a series of incremental updates of a probabilistic model, starting with the model encoding an uninformative prior over admissible solutions and ending with the model that generates only the global optima. (from Wikipedia)

What is One-Max?

The OneMax Problem or BitCounting is a simple problem consisting in maximizing the number of ones of a bitstring.

How is Incremental Learning different from simple EDAs?

In incremental learning approach, instead of generating probability vector in every iteration, we modify the last P as following:

p[i] = p[i] + landa * (best[i] - p[i])

Where landa is the learning rate (i.e. 0.02) and best is the fittest vector in our population.

Executing program

You can run the notebook on Google Colab.

Author