Simple method of resampling an input using diff tracking

gen_and_resample.py

import pathlib
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt


file_name = pathlib.Path("generated.csv")


def generate(min_value: float, max_value: float, normal_dy: float, num_timesteps, axis) -> None:
    sample = np.random.randn()
    total_range = max_value - min_value    
    difference = sample * total_range
    last_value = min_value + difference

    with open(str(file_name), "w", encoding="utf-8") as out_fp:
        out_fp.write(f"{last_value}\n")

        for i in range(num_timesteps):
            sample = np.random.randn()
            difference = sample * normal_dy

            spikeprob = np.random.randn()
            is_spike = spikeprob < 0.01
            if is_spike:
                # let's amp up the signal 
                difference *= 2
            
            last_value += difference
            out_fp.write(f"{last_value}\n")

    df = pd.read_csv(str(file_name), header=None, names=["memory"])
    print(df.head())

    axis.set_title("original faux data")
    axis.set_ylabel("memory consumption")
    axis.set_xlabel("time")
    axis.plot(df, linewidth=2, color="magenta", label="original")


def generate_sin(xmin: float, xmax: float, axis) -> None:
    x = np.linspace(xmin, xmax, 100000)
    y = np.sin(x)

    with open(str(file_name), "w", encoding="utf-8") as out_fp:
        for yval in y:
            out_fp.write(f"{yval}\n")

    df = pd.read_csv(str(file_name), header=None, names=["memory"])
    print(df.head())

    axis.set_title(f"original faux data, sz={df.shape[0]}")
    axis.set_ylabel("memory consumption")
    axis.set_xlabel("time")
    axis.plot(df, linewidth=2, color="magenta", label="original")



def downsample(factor: float, axis) -> None:
    df = pd.read_csv(str(file_name), header=None, names=["memory"])
    threshold = (df.max() - df.min()).memory / factor

    value = df.iloc[0].memory

    resampled_file = str(file_name).replace(".", f"_resampled_{factor}.")
    with open(resampled_file, "w", encoding="utf-8") as out_fp:
        for row in range(df.shape[0]):
            curr_value = df.iloc[row].memory

            if abs(abs(curr_value) - abs(value)) > threshold:
                out_fp.write(f"{row},{curr_value}\n")
                value = curr_value

    df2 = pd.read_csv(resampled_file, index_col=0)

    axis.set_title(f"RESAMPLED f={factor},sz={df2.shape[0]}")
    axis.set_ylabel("memory consumption")
    axis.set_xlabel("time")
    axis.plot(df2, linewidth=1, alpha=0.8, color="cyan", label="resampled")
    
if __name__ == "__main__":
    fig, ((ax1, ax2, ax3), (ax4, ax5, ax6), (ax7, ax8, ax9)) = plt.subplots(3, 3)
    fig.set_figheight(12)
    fig.set_figwidth(12)

    generate(1000, 1500, 10, 100_000, ax1)
    # generate_sin(-10, 10, ax1)

    axes = [ax2, ax3, ax4, ax5, ax6, ax7, ax8, ax9]

    for axis, factor in zip(axes, (5, 7, 10, 20, 50, 100, 150, 200)):
        downsample(factor, axis)

    plt.savefig(pathlib.Path(file_name).with_suffix(".png"))
    

Oscillating data