Jupyter Notebook交互式降噪绘图：滑块联动逻辑异常求助

I get it—having the noise regenerate every time you adjust the smoothing slider is super frustrating, especially when you just want to see how different sigma values affect the same noisy signal. Let's fix this by adding a simple state management layer to your code, so noise only updates when the counts slider changes.

Here's a clean, memory-efficient solution using a class to track the current noise state:

import ipywidgets as widgets
import matplotlib.pyplot as plt
import numpy as np
import skimage
%matplotlib inline

class NoiseFilterVisualizer:
    def __init__(self, base_signal):
        self.base_signal = base_signal
        self.current_counts = None
        self.current_noise = None
    
    def _update_noise_if_needed(self, counts):
        # Only regenerate noise if the counts value has changed
        if counts != self.current_counts:
            normalized_signal = self.base_signal / np.sum(self.base_signal) * counts
            self.current_noise = np.random.poisson(normalized_signal)
            self.current_counts = counts
        return self.current_noise
    
    def plot(self, counts, sigma):
        noise = self._update_noise_if_needed(counts)
        filtered = skimage.filters.gaussian(noise, sigma)
        
        # Clear previous plot to avoid overlapping lines
        plt.clf()
        plt.plot(noise / np.max(noise), label='Noisy Signal')
        plt.plot(filtered / np.max(filtered), label='Filtered Signal')
        plt.legend()
        plt.xlabel('Sample Index')
        plt.ylabel('Normalized Amplitude')
        plt.show()

# Initialize our visualizer with the base signal
base_sig = np.heaviside(np.linspace(-1,1,100),1)+1
visualizer = NoiseFilterVisualizer(base_sig)

# Create sliders
counts_slider = widgets.IntSlider(min=100, max=10000, step=10, description='Counts')
sigma_slider = widgets.IntSlider(min=1, max=100, description='Smoothing')

# Hook up the interaction
widgets.interact(visualizer.plot, counts=counts_slider, sigma=sigma_slider)

How this works: #

• The NoiseFilterVisualizer class keeps track of the last used counts value and the corresponding noise array in current_counts and current_noise.
• The _update_noise_if_needed method checks if the incoming counts value is different from the stored one. If it is, it regenerates the noise; otherwise, it returns the existing noise array.
• We use plt.clf() to clear the previous plot before drawing new lines, which prevents messy overlapping plots as you adjust sliders.

This approach is far more memory-efficient than pre-storing multiple noise arrays—we only keep one noise array in memory at a time, and it only updates when you actually change the noise level. It should work perfectly with your conda environment versions (Python 3.7.3, ipywidgets 7.5.1, etc.).

内容的提问来源于stack exchange，提问作者Dhirschm