fffchameleon / ec_hw_2024

EC_hw_2024

If you encounter any issues or have any questions regarding this repository, please don't hesitate to open an issue!

Latest Announcement

1. (4/6) judger.py has some error when execute python code, already fixed it.
2. (4/3) Update judger.py and cpp/Makefile

Archived Announcement

(3/27) This is the first time we are grading in this manner. If this method of code evaluation truly reflects accuracy, we will continue to use this approach for assessing implementations in the future. Therefore, if you have any questions or encounter any issues, please don't hesitate to ask. Thank you!

(3/27) When you run judger, please set 0.2 as the alpha of whole arithmetic and set 0.5 as the probability of uniform crossover. But you can test other values as your experiment and write the result and analysis in your report.

(3/25) The testcase/*.out is released, and here's a typo in 09.in and judger.py, please pull the new version using:

git pull origin main

(3/19) The release of testcase/*.out files will be delayed. Expected upload date: March 24.

Sample Code Usage

Clone this repository to your local machine,

git clone https://github.com/fffchameleon/EC_hw_2024.git
cd EC_hw_2024

If you choose to write this assignment in C++,

cd cpp/ && make
./main -n 10 -r binary -p 100 -u 0 -c 0.9 -m 0.1 -g 500 -d

If you prefer to use Python, use,

cd py
python3 main.py -n 10 -r binary -p 100 -u 0 -c 0.9 -m 0.1 -g 500 -d

Both commands should output the following information,

-------------------------------------------
|Parameter           |Value               |
-------------------------------------------
|dimension           |10                  |
|representation      |binary              |
|population_size     |100                 |
|uniform_crossover   |false               |
|crossover_method    |2-point             |
|cross_prob          |0.9                 |
|mut_prob            |0.1                 |
|num_generations     |500                 |
-------------------------------------------
0.00145984

Input/Output Format

We provide sample parser code for two languages (C++/Python).

You can write your own parser, but it must be capable of accepting the following parameters, and your program must provide at least the following 8 options.

p.s. You may add more options to make your experiments more convenient and complete.

Options	Description	Default
-n, --dimension	The dimension of Schwefel function	10
-r, --representation	The representation to use. Binary or real-valued (binary, real)	binary
-p, --population_size	Number of the population	100
-u, --uniform_crossover	The crossover method using uniform crossover (1) or not (0). If not, then for binary GA, it will use 2-point crossover and for real-valued GA will use whole arithmetic crossover	0
-c, --pc	Probability for the crossover	$p_c$=0.9
-m, --pm	Probability for the mutation	$p_m$=0.1
-g, --generations	Max number of generations to terminate	500
-d, --debug	Turn on debug prints	false

Input

For example, with testcase/01.in, our judger will execute your executable file + the first line of configuration in testcase/*.in.

If written in C++, please remember to upload a Makefile, and the compiled executable should be named main.

./main -n 10 -r binary -p 100 -u 0 -c 0.9 -m 0.1 -g 500

If written in Python, your main.py file will be executed directly.

python3 ./main.py -n 10 -r binary -p 100 -u 0 -c 0.9 -m 0.1 -g 500

The following context is 01.in.

1. The first line contains the parameters we will use for testing
2. The second line is a configuration table. If you do not use -d or --debug, it will not be printed.

-n 10 -r binary -p 100 -u 0 -c 0.9 -m 0.1 -g 500
-------------------------------------------
|Parameter           |Value               |
-------------------------------------------
|dimension           |10                  |
|representation      |binary              |
|population_size     |100                 |
|uniform_crossover   |false               |
|crossover_method    |2-point             |
|cross_prob          |0.9                 |
|mut_prob            |0.1                 |
|num_generations     |500                 |
-------------------------------------------

Output

Remember, when uploading your code, ensure that your code only print the configuration table when -d, --debug is added.

Please directly print the best fitness of the final generation.

For example, with 01.in, suppose I finally find the best solution 430 420 419 432 426 419 415 417 414 429, and the fitness of my solution is 50.5991. Then just print 50.5991.

In testcase/*.out, each file contains two lines: the first line is the mean best fitness running by TAs with 120 trials, and the second line is the standard deviation.

We will run your code for 30 trials and if your mean fitness is within two standard deviations, it is considered correct.

Judger Usuage

Place your executable and the judger in the same directory, and run them using python3 judger.py.

It may take a long time to run judge (each testcse will run 30 trials), so be patient!

Grading

• Coding (30%)

  • Simple test: each failed testcase deducts 1 point (10%)
  • Parent Selection (3%)
  • Crossover (8%)
  • Mutation (6%)
  • Survivor Selection (3%)
  • Penalty
    • Compilation failure with Makefile (-5)
    • Not following input/output format (-5)
    • Incorrect filename (-3)
      • executable file: main or main.py
      • report: report.pdf

• Report (70%)