|
|
import gradio as gr |
|
|
import numpy as np |
|
|
import matplotlib.pyplot as plt |
|
|
|
|
|
|
|
|
|
|
|
def run_snn(X, epochs, batch_size, l2_lambda, patience): |
|
|
|
|
|
snn = SwarmNeuralNetwork(layer_sizes=[1, 32, 16, 8, 1], |
|
|
fractal_methods=[sierpinski_fractal, mandelbrot_fractal, julia_fractal, julia_fractal]) |
|
|
snn.train(X, y, epochs=epochs, batch_size=batch_size, l2_lambda=l2_lambda, patience=patience) |
|
|
y_pred = snn.forward(X, training=False) |
|
|
fractal_outputs = snn.apply_fractals(X) |
|
|
return y_pred, fractal_outputs |
|
|
|
|
|
def plot_results(X, y, y_pred, fractal_outputs): |
|
|
fig, axs = plt.subplots(2, 2, figsize=(15, 10)) |
|
|
|
|
|
axs[0, 0].plot(X, y, label='True') |
|
|
axs[0, 0].plot(X, y_pred, label='Predicted') |
|
|
axs[0, 0].legend() |
|
|
axs[0, 0].set_title('True vs Predicted') |
|
|
|
|
|
x, y = fractal_outputs[0] |
|
|
axs[0, 1].plot(x, y) |
|
|
axs[0, 1].set_title('Sierpinski Fractal Output') |
|
|
|
|
|
axs[1, 0].plot(X, fractal_outputs[1]) |
|
|
axs[1, 0].set_title('Mandelbrot Fractal Output') |
|
|
|
|
|
axs[1, 1].plot(X, fractal_outputs[2]) |
|
|
axs[1, 1].set_title('Julia Fractal Output') |
|
|
|
|
|
return fig |
|
|
|
|
|
def main_interface(epochs, batch_size, l2_lambda, patience): |
|
|
X = np.linspace(0, 10, 1000).reshape(-1, 1) |
|
|
y = np.sin(X).reshape(-1, 1) |
|
|
X = (X - X.min()) / (X.max() - X.min()) |
|
|
y = (y - y.min()) / (y.max() - y.min()) |
|
|
|
|
|
y_pred, fractal_outputs = run_snn(X, epochs, batch_size, l2_lambda, patience) |
|
|
fig = plot_results(X, y, y_pred, fractal_outputs) |
|
|
|
|
|
return fig |
|
|
|
|
|
gr.Interface( |
|
|
fn=main_interface, |
|
|
inputs=[ |
|
|
gr.inputs.Slider(1, 10000, default=5000, label="Epochs"), |
|
|
gr.inputs.Slider(1, 100, default=32, label="Batch Size"), |
|
|
gr.inputs.Slider(0.0001, 0.1, default=0.00001, label="L2 Lambda"), |
|
|
gr.inputs.Slider(1, 1000, default=50, label="Patience") |
|
|
], |
|
|
outputs=gr.outputs.Plot() |
|
|
).launch() |
|
|
|
