Hugging Face's logo

Spaces:
SLU-CSCI4750
/
gradient_descent
Build error

App Files Community
gradient_descent
File size: 4,487 Bytes
a38e034
 
 
 
7565946
 
a38e034
 
 
 
792c809
a38e034
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
97ef1c3
792c809
a38e034
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
792c809
 
 
ee01add
7565946
792c809
 
 
 
 
a38e034
 
 
792c809
556b56a
 
 
 
a38e034
556b56a
 
 
a38e034
556b56a
a38e034
7565946
a38e034
 
 
 
 
 
 
 
 
b128848
a38e034
 
 
033b3d6
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
def homework04_solution(theta0, theta1, theta2, learning_rate):
  import numpy as np
  import pandas as pd

  print(theta0, theta1, theta2, learning_rate)

  def linear_predict(b0, b1, b2, x1, x2):
    y_hat = b0 + b1*x1 + b2*x2
    return y_hat

    
  def get_linear_results(data, theta0, theta1, theta2):
    ## (2) make linear prediction
    y_hat_list = []

    theta0_grad = 0
    theta1_grad = 0
    theta2_grad = 0
    for i in range(len(data)):
      x1 = data.iloc[i,0]
      x2 = data.iloc[i,1]
      y = data.iloc[i,2]
      y_hat = linear_predict(theta0, theta1, theta2, x1, x2)
      y_hat_list.append(y_hat)

      ## (3) calculate gradients
      theta0_grad = theta0_grad - 2/len(data)*(y - theta0 - theta1*x1 - theta2*x2)*1.0

      theta1_grad = theta1_grad - 2/len(data)*(y - theta0 - theta1*x1 - theta2*x2)*x1

      theta2_grad = theta2_grad - 2/len(data)*(y - theta0 - theta1*x1 - theta2*x2)*x2

    data['y_hat'] = y_hat_list

    data['y-y_hat'] = data['y'] - data['y_hat']
    data['(y-y_hat)^2'] = data['y-y_hat']*data['y-y_hat']
    return data, theta0_grad, theta1_grad, theta2_grad

  ## (1) load data
  X = np.array([[15,20], [30,16], [12,6.5], [13,20], [18,18]])
  y = [4.9, 5.8,6.5,7.3,7.2]
  data = pd.DataFrame(X, columns=['X1','X2'])
  data['y'] = y 

  ## (2) get regression table, gradients
  data, theta0_grad, theta1_grad, theta2_grad = get_linear_results(data, theta0, theta1, theta2)


  ### (3) summarize gradient results for question 3a

  data_t = data.T

  data_t = data_t.round(2)
  
  data_t.insert(loc=0, column='Name', value=['X1', 'X2', 'y', 'y_hat', 'y-y_hat', '(y-y_hat)^2'])

  data_t.columns = ['Name', '0', '1', '2', '3', '4']
  

  ### (4) summarize gradient results for question 3b

  MSE = data['(y-y_hat)^2'].mean()

  q3_mse  = MSE
  
  ### summarize gradient results for question 4 (2)

  ### update parameter using gradient descent  4 (3)
  
  theta0_new = theta0 - learning_rate*theta0_grad
  theta1_new = theta1 - learning_rate*theta1_grad
  theta2_new = theta2 - learning_rate*theta2_grad


  ### (5) recalculate linear regression table using new gradients
  data4,_,_,_ = get_linear_results(data, theta0_new, theta1_new, theta2_new)


  ### (6) summarize gradient results for question 4 (4)

  MSE = data4['(y-y_hat)^2'].mean()

  q4_mse  = MSE
  
  ### (7) return all results for Gradio visualization
  return data_t, q3_mse, theta0_grad, theta1_grad , theta2_grad, theta0_new, theta1_new, theta2_new, q4_mse

import numpy as np

import gradio as gr


### configure inputs
set_theta0 = gr.Number(value=0.1)
set_theta1 = gr.Number(value=0.1)
set_theta2 = gr.Number(value=0.1)
set_ita = gr.Number(value=0.001)

#print("check8")
#set_theta0 = gr.Slider(minimum = -10, maximum=10, step=0.01, value=0.1)
#set_theta1 = gr.Slider(minimum = -10, maximum=10, step=0.01, value=0.1)
#set_theta2 = gr.Slider(minimum = -10, maximum=10, step=0.01, value=0.1)
#set_ita = gr.Slider(minimum = 0, maximum=10, step=0.01, value=0.01)


### configure outputs
set_output_q3a = gr.Dataframe(type='pandas', label ='Question 3a')
set_output_q3b = gr.Textbox(label ='Question: What\'s Initial MSE loss')
set_output_q4a0 = gr.Textbox(label ='Question: What\'s theta0_grad')
set_output_q4a1 = gr.Textbox(label ='Question: What\'s theta1_grad')
set_output_q4a2 = gr.Textbox(label ='Question: What\'s theta2_grad')

set_output_q4b0 = gr.Textbox(label ='Question: What\'s theta0_new: updated by gradient descent')
set_output_q4b1 = gr.Textbox(label ='Question: What\'s theta1_new: updated by gradient descent')
set_output_q4b2 = gr.Textbox(label ='Question: What\'s theta2_new: updated by gradient descent')

set_output_q4b4 = gr.Textbox(label ='Question: What\'s New MSE after update the parameters using gradient descent')


### configure Gradio
interface = gr.Interface(fn=homework04_solution, 
                         inputs=[set_theta0, set_theta1, set_theta2, set_ita], 
                         outputs=[set_output_q3a, set_output_q3b, 
                                  set_output_q4a0, set_output_q4a1, set_output_q4a2, 
                                  set_output_q4b0, set_output_q4b1, set_output_q4b2,
                                  set_output_q4b4],
                         title="CSCI4750/5750(gradient descent): Linear Regression/Optimization", 
                         description= "Click examples below for a quick demo",
                         theme = 'huggingface'
                         )


interface.launch(debug=True)