integraali10 / probalisticalzheimer

Файлы с потаповского курса вероятностного программирования (Church⛪, Racket🎾)

ProbalisticAlzheimer

General Epic To-Do

• Generation
  • Randomise Math Expression with only Numbers using Binary Tree and Polish Notation
  • Randomise Math Expression with One Variable (e.g. x) using Binary Tree and Polish Notation
  • Randomise Math Expression with Two Variables (e.g. x1 (predecessor) and x2(pred-predecessor)) and their iterator n using Binary Tree and Polish Notation to produce Recursive Expression
• Auxiliary Functions
  • Counting Leaves (all numbers and variables)
  • Counting Nodes (all operators (operations))
• Interpretation
  • Interpretation of the first kind of Expression
  • Interpretation of the second kind of Expression with input for x
  • Interpretation of the third kind of Expression with input for Two Variables (e.g. x1 and x2) and their iterator n
• Producing sequences
  • Using Math Expression with one Variable (e.g. x) to produce iterable sequences with length of input N
  • Using Math Expression with Two Variables to produce iterable recursive sequence with length of input N
• Simple Rejection
  • for given sequence predict accuratly Exprerssion with one variable using random
  • for given sequence predict accuratly Recursive Expression with Two Variables using random
• Genetic programming for given sequence predict accuratly Recursive Expression with Two Variables using Genetic Programming
  • Mutation of Expression
    • Changing operators and some of nums/vars but saving structure
    • Completely change on of the branches in random way
  • Crossbreeding two Expressions (Realization on only one of rules is permissible)
    • Change random branch of parent1 to random branch of parent2
    • Change subtree of parent1 to subtree of parent2
  • Parents Selection
    • distant relationship
    • closest relationship
    • random selection
  • Fitness Function (who is the best to be producer, or to be a candidate) - Mean Squared Error (MSE)
  • Managing the population (love, death and KPI)
  • Benchmarks - time consuming of simple rejection and genetic one.

Specific To-do

• trees_sequences_actual.rkt
  • finish recursive sequences generation and prediction
• trees_sequence_genetic_generation.rkt - Potapov's solution + mutation
  • add crossbreeding (see rules in orangebook)
  • add parents selection rule
  • add population structure
  • add fitness function - MSE