quotegp
所属分类：人工智能/神经网络/深度学习
开发工具：Elixir
文件大小：10KB
下载次数：0
上传日期：2022-09-15 13:34:58
上 传 者：
sh-1993
说明： 使用Elixir AST的遗传编程
(Genetic programming using the Elixir AST)
文件列表:
.formatter.exs (97, 2022-09-15)
lib (0, 2022-09-15)
lib\quotegp.ex (25, 2022-09-15)
lib\quotegp (0, 2022-09-15)
lib\quotegp\cli.ex (452, 2022-09-15)
lib\quotegp\code.ex (942, 2022-09-15)
lib\quotegp\config.ex (547, 2022-09-15)
lib\quotegp\generation.ex (2586, 2022-09-15)
lib\quotegp\genetic_operators.ex (2091, 2022-09-15)
lib\quotegp\population.ex (3218, 2022-09-15)
mix.exs (630, 2022-09-15)
mix.lock (2167, 2022-09-15)
samples (0, 2022-09-15)
samples\sample1.json (434, 2022-09-15)
samples\sample2.json (529, 2022-09-15)
samples\sample3.json (660, 2022-09-15)
test (0, 2022-09-15)
test\quotegp_test.exs (2825, 2022-09-15)
test\test_helper.exs (15, 2022-09-15)
# QuoteGP - genetic programming with native Elixir code
A genetic programming system that uses Elixir AST as a program representation and can evolve native Elixir language
solutions to problems.
This is currently at a proof-of-concept level of maturity.
## Getting Started & Samples
There are a couple of simple sample problems included in this repo, in the
`samples` subdirectory. The problems here are "symbolic regression" curve fitting
problems, in which the GP system attempts to evolve code which maps a series of
numerical inputs to the provided outputs.
These samples are "toy" problems - they are intended to be solvable but slightly
challenging in order to test the GP system and demonstrate how it works.
To run via the command-line, you can use the included escript:
```
mix escript.build
./quotegp ./samples/sample1.json
```
To run via code:
```
# Generate some test data:
test_cases = 8..16 |> Enum.map(fn x -> {x, x * x + 3 * x + 8} end)
# Configure and run:
config = %QuoteGP.Config{}
QuoteGP.Population.build(config)
|> QuoteGP.Population.run(test_cases)
```
For more advanced usage, configure a GP run with the following options (shown with defaults):
```
%QuoteGP.Config{
population_size: 1000,
max_generations: 100,
max_program_depth: 10,
mutation_probability: 0.1,
halt_fitness: 0.0,
mutation_rate: 0.3,
crossover_rate: 0.3,
tournament_size: 9
}
```
- halt_fitness: a fitness at which execution will halt
- max_generations: the maximum number of generations to run if a solution is not found
- population_size: the number of individuals in the evolving population
- max_program_depth: the maximum depth of a program in the population
- mutation_probability: 0.1,
- mutation_rate: 0.3,
- crossover_rate: 0.3,
- tournament_size: the number of individuals in a "tournament" selection event
## Other Classes of Problems
近期下载者：
相关文件：
收藏者：