app.py

### CSCI 4750/5750- SLU
### Jie Hou, Oct 2022
import os
os.system("pip uninstall -y gradio")
os.system("pip install gradio==2.6.4")

import gradio as gr
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
import matplotlib.colors as colors
import itertools
from scipy.stats import norm
from scipy import stats
from sklearn.naive_bayes import GaussianNB



def gaussian(x, n, u, s):
    #u = x.mean()
    #s = x.std()

    # divide [x.min(), x.max()] by n
    x = np.linspace(x.min(), x.max(), n)

    a = ((x - u) ** 2) / (2 * (s ** 2))
    y = 1 / (s * np.sqrt(2 * np.pi)) * np.exp(-a)

    return x, y, u, s
    
import gradio as gr
# 1. define mean and standard deviation for class 1

set_fea1_mean_class1 = gr.inputs.Slider(0, 20, step=0.5, default=1, label = 'Feature_1 Mean (Class 1)')
set_fea1_var_class1 = gr.inputs.Slider(0, 10, step=0.5, default=2.5, label = 'Feature_1 Variance (Class 1)')

set_fea2_mean_class1 = gr.inputs.Slider(0, 20, step=0.5, default=2, label = 'Feature_2 Mean (Class 1)')
set_fea2_var_class1 = gr.inputs.Slider(0, 10, step=0.5, default=4.5, label = 'Feature_2 Variance (Class 1)')

set_fea_covariance_class1 = gr.inputs.Slider(0, 10, step=0.5, default=2.5, label = 'Feature_1_2 Co-Variance (Class 1)')

# 2. define mean and standard deviation for class 2

set_fea1_mean_class2 = gr.inputs.Slider(0, 20, step=0.5, default=7, label = 'Feature_1 Mean (Class 2)')
set_fea1_var_class2 = gr.inputs.Slider(0, 10, step=0.5, default=3.5, label = 'Feature_1 Variance (Class 2)')

set_fea2_mean_class2 = gr.inputs.Slider(0, 20, step=0.5, default=8, label = 'Feature_2 Mean (Class 2)')
set_fea2_var_class2 = gr.inputs.Slider(0, 10, step=0.5, default=3.5, label = 'Feature_2 Variance (Class 2)')

set_fea_covariance_class2 = gr.inputs.Slider(0, 10, step=0.5, default=3.5, label = 'Feature_1_2 Co-Variance (Class 2)')

# 3. Define the number of data points 
set_number_points = gr.inputs.Slider(10, 100, step=5, default=20, label = 'Number of samples in each class')

# 5. set classifier type
set_classifier = gr.inputs.Dropdown(["None", "LDA", "QDA", "NaiveBayes"])

# 6. define output imagem model
set_out_plot_images = gr.outputs.Image(label="Data visualization")

set_out_plot_table = gr.outputs.Dataframe(type='pandas', label ='Simulated Dataset')



def plot_figure_twofeature(N, fea1_u1, fea1_var1, fea2_u1, fea2_var1, covariance1, fea1_u2, fea1_var2, fea2_u2, fea2_var2, covariance2, classifier=None):
  
  
  #N = 100
  import numpy as np
  import matplotlib.pyplot as pp
  pp.style.use('default')
  val = 0. # this is the value where you want the data to appear on the y-axis.

  np.random.seed(seed = 3)

  mu1 = [fea1_u1, fea2_u1]
  sigma1 = [[np.sqrt(fea1_var1), np.sqrt(covariance1)], [np.sqrt(covariance1), np.sqrt(fea2_var1)]]
  points_class1_fea1, points_class1_fea2 = np.random.multivariate_normal(mu1, sigma1, N).T

  mu2 = [fea1_u2, fea2_u2]
  sigma2 = [[np.sqrt(fea1_var2), np.sqrt(covariance2)], [np.sqrt(covariance2), np.sqrt(fea2_var2)]]
  points_class2_fea1, points_class2_fea2 = np.random.multivariate_normal(mu2, sigma2, N).T

  mu_list = [mu1,mu2]
  sigma_list = [sigma1,sigma2]
  color_list = ['darkblue','darkgreen']


  pd_class1 = pd.DataFrame({'Feature 1 (X)': points_class1_fea1,'Feature 2 (X)': points_class1_fea2, 'Label (Y)': np.repeat(0,len(points_class1_fea1))})
  pd_class2 = pd.DataFrame({'Feature 1 (X)': points_class2_fea1,'Feature 2 (X)': points_class2_fea2, 'Label (Y)': np.repeat(1,len(points_class2_fea1))})


  pd_all = pd.concat([pd_class1, pd_class2]).reset_index(drop=True)

  import numpy as np
  #X_data= pd_all['Feature 1 (X)','Feature 2 (X)'].to_numpy().reshape((len(pd_all),2))
  #y_labels= pd_all['Label (Y)']


  #Setup X and y data
  X_data = np.asarray(np.vstack((np.hstack((points_class1_fea1, points_class1_fea2)),np.hstack((points_class2_fea1, points_class2_fea2)))).T)
  y_labels = np.hstack((np.zeros(N),np.ones(N)))

  print("X_data: ",X_data.shape)

  fig = pp.figure(figsize=(8, 6)) # figure size in inches
  fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.3, wspace=0.05) 


  #pp.tick_params(left = False, right = False , labelleft = True ,
  #                labelbottom = True, bottom = False)

  #reference = [stats.uniform.rvs(loc=1, scale = 1) for x in range(N)]
  pp.plot(points_class1_fea1, points_class1_fea2 + val, 'x', label = 'Class 1', markersize = 10)
  pp.plot(points_class2_fea1, points_class2_fea2 + val, 'o', label = 'Class 2', markersize = 10)



  
  # define x, y limits
  #x_min, x_max = X_data[:, 0].min() - 1, X_data[:, 0].max() + 1
  #y_min, y_max = X_data[:, 1].min() - 1, X_data[:, 1].max() + 1

  x_min, x_max = -5, 15
  y_min, y_max = -5, 15
  X_grid, Y_grid = np.meshgrid(np.linspace(x_min, x_max, 100),
                          np.linspace(y_min, y_max, 100))
  


  ### draw decision boundary

  import numpy as np
  from matplotlib import pyplot as plt
  from sklearn import neighbors, datasets
  from matplotlib.colors import ListedColormap

  # Create color maps for 3-class classification problem, as with iris
  cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA'])
  cmap_bold = ListedColormap(['#FF0000', '#00FF00'])


  if classifier == 'LDA':
    from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
    model_sk = LinearDiscriminantAnalysis()
    model_sk.fit(X_data,y_labels)

    X_grid, Y_grid = np.meshgrid(np.linspace(x_min, x_max, N),
                          np.linspace(y_min, y_max, N))
    #Predictions for each point on meshgrid
    zz = np.array(  [model_sk.predict( [[xx,yy]])[0] for xx, yy in zip(np.ravel(X_grid), np.ravel(Y_grid)) ] )

    Z = zz.reshape(X_grid.shape)

    pp.pcolormesh(X_grid, Y_grid, Z, cmap=cmap_light, alpha=0.2)
    #pp.contour( X_grid, Y_grid, Z, 1, alpha = .3, colors = ('red'))

  elif classifier == 'QDA':
    from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
    model_sk = QuadraticDiscriminantAnalysis()
    model_sk.fit(X_data,y_labels)

    X_grid, Y_grid = np.meshgrid(np.linspace(x_min, x_max, N),
                          np.linspace(y_min, y_max, N))
    #Predictions for each point on meshgrid
    zz = np.array(  [model_sk.predict( [[xx,yy]])[0] for xx, yy in zip(np.ravel(X_grid), np.ravel(Y_grid)) ] )

    Z = zz.reshape(X_grid.shape)

    pp.pcolormesh(X_grid, Y_grid, Z, cmap=cmap_light, alpha=0.2)

  elif classifier == 'NaiveBayes':
    from sklearn.naive_bayes import GaussianNB
    model_sk = GaussianNB(priors = None)
    model_sk.fit(X_data,y_labels)

    X_grid, Y_grid = np.meshgrid(np.linspace(x_min, x_max, N),
                          np.linspace(y_min, y_max, N))
    #Predictions for each point on meshgrid
    zz = np.array(  [model_sk.predict( [[xx,yy]])[0] for xx, yy in zip(np.ravel(X_grid), np.ravel(Y_grid)) ] )

    Z = zz.reshape(X_grid.shape)

    pp.pcolormesh(X_grid, Y_grid, Z, cmap=cmap_light, alpha=0.2)


  #Reshaping the predicted class into the meshgrid shape

  #Plot the  contours
  


  print(x_min, x_max)
  print(y_min, y_max)
  #pp.xlim([x_min-0.8*x_max, x_max+0.8*x_max])
  #pp.ylim([y_min-0.8*y_max, y_max+0.8*y_max])
  pp.xlim([x_min, x_max])
  pp.ylim([y_min, y_max])
  pp.xlabel("Feature 1 (X)", size=20)
  pp.xticks(fontsize=20)
  pp.yticks(fontsize=20)
  pp.ylabel("Feature 2 (X)", size=20)
  pp.legend(loc='upper right', borderpad=0, handletextpad=0, fontsize = 20)
  pp.savefig('plot.png')

  return 'plot.png', pd_all


### configure gradio, detailed can be found at https://www.gradio.app/docs/#i_slider
interface = gr.Interface(fn=plot_figure_twofeature, inputs=[set_number_points,set_fea1_mean_class1,set_fea1_var_class1,set_fea2_mean_class1,set_fea2_var_class1,set_fea_covariance_class1,set_fea1_mean_class2,set_fea1_var_class2,set_fea2_mean_class2,set_fea2_var_class2,set_fea_covariance_class2, set_classifier], 
                         outputs=[set_out_plot_images,set_out_plot_table],
                         examples_per_page = 2,
                         #examples = get_sample_data(10), 
                         title="CSCI4750/5750 Demo: Web Application for Probabilistic Classifier (Two features)", 
                         description= "Click examples below for a quick demo",
                         theme = 'huggingface'
                         )
interface.launch(debug=True)