https://github.com/skycoin/cx-evolves
https://github.com/skycoin/cx-evolves
Last synced: 4 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/skycoin/cx-evolves
- Owner: skycoin
- Created: 2020-10-08T03:01:04.000Z (over 5 years ago)
- Default Branch: develop
- Last Pushed: 2021-10-25T04:28:39.000Z (over 4 years ago)
- Last Synced: 2024-06-21T15:32:20.862Z (almost 2 years ago)
- Language: Go
- Size: 2.95 MB
- Stars: 6
- Watchers: 7
- Forks: 2
- Open Issues: 22
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Installation
- Install and setup go 1.14 and above.
- Install CX (https://github.com/skycoin/cx).
- `git pull` your CX repository so you have the latest CX version.
- Make sure `go module` is enabled. Install go dependencies by running:
```
$go mod verify
$go mod tidy
$go mod download
```
# Benchmarks
| Task Name | Task Code Name | Task Version |
| ---------------------- |:-------------------:|:------------:|
| Maze | "maze" | 1 |
| Constants | "constants" | 1 |
| Evens | "evens" | 1 |
| Evens greater than 0 | "evens" | 2 |
| Odds | "odds" | 1 |
| Primes | "primes" | 1 |
| Composites | "composites" | 1 |
| Network Simulator | "network_simulator" | 1 |
| Range | "range" | 1 |
### Summary
This program runs evolve with Maze. Generated programs are given multiple i32 as inputs and outputs one i32. The output is their move for the maze.
### Usage
For more information, run
```
go run main.go help
```
```
go run main.go
--task=[Name of task to benchmark (in lower case--refer to task code name above)]
--task-version=[Task vesion value, default is 1]
--population=[Population size or number of programs generated per generation]
--generations=[number of generations]
--expressions=[number of expressions a generated program can have]
--graphs=[Set true if average fitness and fittest per generation graphs will be saved]
--ast=[Set true if best ASTs per generation will be saved]
--use-log-fitness=[Set true if fitness will be fitness(log base 2)]
--workers=[number of workers available to use, if not set default is 1]
--random-search=[set true to have no mutation on individuals]
--select-rank-cufoff=[set true if selection is select, rank, and cutoff]
-W=[width of the maze to solve]
-H=[height of maze to solve]
--random-maze-size=[set true if generated maze for every N epochs will have random sizes (NxN 2,3,4,5,6,7, or 8)]
--epoch-length=[Maze changes for every N generation(Example. Epochs=5, maze changes every 5 generations)]
--upper-range=[upper range (int) for range benchmark]
--lower-range=[lower range (int) for range benchmark]
--rounds=[Number of rounds per program]
--constants-target=[target number for constants benchmark]
```
### Example Benchmarking
For Maze
```
go run main.go --task=maze -W=2 -H=2 --random-maze-size=false --population=50 --generations=100 --expressions=30 --epoch-length=100 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Range
```
go run main.go --task=range --upper-range=9 --lower-range=2 --rounds=10 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Constants
```
go run main.go --task=constants --rounds=10 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Constants with explicit target
```
go run main.go --task=constants --rounds=10 --constants-target=5 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
Note: For constants benchmarking 0-256 range, use --constants-target=256.
```
For Odds
```
go run main.go --task=odds --rounds=10 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Evens
```
go run main.go --task=evens --rounds=10 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Primes
```
go run main.go --tasks=primes --rounds=10 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Composites
```
go run main.go --task=composites --rounds=10 --population=300 --generations=1000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
For Network Simulator
```
go run main.go --task=network_simulator --rounds=20 --population=300 --generations=2000 --expressions=30 --graphs=true --ast=false --use-log-fitness=false --workers=1
```
### Notes
1. If no arguments are specified, the program will run default values.
# Benchmarking in Docker Env
First edit the ./scripts/maze_benchmark.sh to the benchmark options needed and the number of workers to deploy.
Then use this command:
To build and run
```
NAME=[Name for the docker image]
MOUNT=[Local directory to mount the "Result" directory of the container, this is where the graphs and asts can be found]
make deploy
```
Example
```
NAME=testbenchmark MOUNT=/Benchmarking/Results make deploy
```
To build and push to dockerhub
```
NAME=[Name for the docker image]
make push-docker
```
Example
```
NAME=testbenchmark make push-docker
```
# High priority tasks
- [x] Maze
- [ ] Snake
- [ ] Tic-Tac-Toe
- [ ] Multi-armed bandit
- [ ] Rock, paper & scissors
# Low priority tasks
- [ ] Time-series mean estimation
- [ ] Time-series variance estimation
- [ ] Time-series covariance estimation
- [ ] Stock market simulator
- [ ] Committee of experts
- [ ] Linear feedback shift register prediction
- [ ] Non-linear feedback shift register prediction
- [x] Check if a number is prime