Research on Adaptive Mutations

[gkirgizov]

The idea is to adapt mutations during optimizations process, so that mutation probabilities would help in faster and better convergence.

Each mutation can be considered as Action in certain State (graph or node context) with certain Reward. This perspective opens several possibilities of multi-armed bandit formulation and RL formulation of adaptive mutations.

Subissues:

• Implement universal & complete set of mutations #39
• Implement infrastructure for adaptive mutations #41
• Experiment with MultiArmedBandit implementation of Adaptive Mutations #42
• Implement state embedding of nodes/graphs #77
• Experiment with RL impl of Adaptive Mutations #43
• Conduct experiments on supervised learning with Adaptive agent #98

Following is also useful:

• Add example of Tree search with Edit Distance metric #38
• Molecule search example #83