analyse.py

import argparse
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd

parser = argparse.ArgumentParser(
    prog="Analyse results", description="Analyse and compare models"
)
parser.add_argument("--path", default="results/genetic_1705104088")
args = parser.parse_args()


def analyse(path=None):
    model_path = path or args.path
    scores = np.loadtxt(f"{model_path}/scores.txt")
    line_history = np.loadtxt(f"{model_path}/line_history.txt")
    print_descriptive(scores, line_history)
    show_cleared_lined_per_step(line_history)


def print_descriptive(scores, line_history):
    print(f"n samples:\t{len(scores)}")
    print(f"Mean:\t{np.mean(scores)}")
    print(f"Std:\t{np.std(scores)}")
    print(f"Min:\t{np.min(scores)}")
    print(f"Quantile:\t{np.quantile(scores, [0.25, 0.50, 0.75])}")
    print(f"Max:\t{np.max(scores)}")
    avg_reward = np.sum(scores) / np.sum(np.sum(line_history))
    print(f"Avg score per move:\t{avg_reward}")


def show_cleared_lined_per_step(line_history):
    sns.set(style="darkgrid")
    df = pd.DataFrame(line_history)
    column_sums = df.sum()
    sns.barplot(x=column_sums.index, y=column_sums.values)
    plt.title("Histogram over lines cleared per step")
    plt.xlabel("Lines cleared")
    plt.ylabel("Number of steps")
    plt.xticks(range(line_history.shape[1]))
    plt.show()


if __name__ == "__main__":
    analyse()