Usage

ez-animate is compatible with both Scikit-learn and sega_learn models, making it a versatile tool for creating animations of machine learning model behavior. See below for examples of how to use ez-animate with different types of models. Or see API Reference for more details on available methods and customization options.

More advanced usage examples can be found in the on GitHub.

Complete Scikit-learn Example

This section demonstrates how to use ez-animate with a Scikit-learn SGD regression model to create an animation of the model's predictions for max_iter values ranging from 1 to 100.

import numpy as np

from ez_animate import RegressionAnimation
from sega_learn.utils import Metrics, make_regression
from sklearn.linear_model import SGDRegressor

# Generate synthetic regression data
X, y = make_regression(n_samples=1000, n_features=1)

# Create the animation using RegressionAnimation
animator = RegressionAnimation(
    model=SGDRegressor,
    X=X,
    y=y,
    test_size=0.25,
    dynamic_parameter="max_iter",
    static_parameters={"fit_intercept": True, "eta0": 0.0005},
    keep_previous=True,
    max_previous=25,
    metric_fn=[
        Metrics.r_squared,
        Metrics.mean_squared_error,
    ],
    plot_metric_progression=True,
    max_metric_subplots=2,
)

# Set up the plot
animator.setup_plot(
    title="SGD Regression Animation",
    xlabel="Feature Coefficient",
    ylabel="Target Value",
    legend_loc="upper left",
    grid=True,
    figsize=(14, 6),
)

# Create and save the animation
iter_range = np.arange(1, 100, 1)
animator.animate(frames=iter_range, interval=150, blit=False, repeat=True)
animator.show()

Complete sega_learn Example

This section demonstrates how to use ez-animate with a time series forecasting model from sega_learn to animate the effect of the alpha parameter in Exponential Moving Average forecasting.

import numpy as np

from ez_animate import ForecastingAnimation
from sega_learn.time_series.moving_average import ExponentialMovingAverage
from sega_learn.utils import Metrics, make_time_series

# Generate a synthetic time series
time_series = make_time_series(
    n_samples=1,
    n_timestamps=300,
    n_features=1,
    trend="linear",
    seasonality="sine",
    seasonality_period=30,
    noise=0.1,
    random_state=1,
)

# Flatten the time series to 1D if it's not already
time_series = time_series.flatten()

# Split into training and testing sets
train_size = int(len(time_series) * 0.8)
train_series, test_series = time_series[:train_size], time_series[train_size:]
forecast_steps = len(test_series)

# Create the animation using ForecastingAnimation
animator = ForecastingAnimation(
    model=ExponentialMovingAverage,
    train_series=train_series,
    test_series=test_series,
    forecast_steps=forecast_steps,
    keep_previous=True,
    dynamic_parameter="alpha",
    metric_fn=[
        Metrics.mean_squared_error,
        Metrics.mean_absolute_error,
    ],
)

# Set up the plot
animator.setup_plot(
    title="Exponential Moving Average Forecast",
    xlabel="Time Step",
    ylabel="Value",
    legend_loc="upper left",
    grid=True,
    figsize=(12, 6),
)

# Create and show the animation
alpha_range = np.arange(0.01, 0.5, 0.01)
animator.animate(frames=alpha_range, interval=150, blit=True, repeat=False)
animator.show()

Specific Features

This section highlights some of the key features of ez-animate and how to use them with the AnimationBase class and its subclasses.

Customizing Figures and Axes

You can customize the plot's title, axis labels, grid, legend location, and figure size using the setup_plot method:

animator.setup_plot(
    title="Custom Animation Title",
    xlabel="X Axis Label",
    ylabel="Y Axis Label",
    legend_loc="upper right",
    grid=True,
    figsize=(10, 5),
)

Saving Animations

After creating an animation, you can save it as a GIF or MP4 using the save method:

animator.save(
    filename="my_animation.gif",  # or "my_animation.mp4"
    writer="pillow",              # or "ffmpeg" for MP4
    fps=10,
    dpi=150,
)

Controlling Animation Speed and Playback

You can control the speed of the animation using the interval parameter in the animate method (milliseconds between frames), and set whether the animation repeats:

animator.animate(frames=range(1, 100), interval=100, blit=True, repeat=False)

Additional Notes

• The AnimationBase class is designed for extensibility. Subclasses like RegressionAnimation, ClassificationAnimation, and ForecastingAnimation implement model-specific logic.
• You can pass additional keyword arguments to customize plot elements or metrics.
• The show() method displays the animation in a Matplotlib window.

See API Reference for more details on available methods and customization options.