Thu Mar 28 23:18:06 2024

montana():
  python version: 3.10.12
  keras version: 3.1.1
  Neural network to solve a 1d regression problem.
  Find weights (a,b) so that the formula
    y = a x + b
  best approximates a set of (x,y) data.
  The data is read from internal functions.
  There are 100 pairs of (x,y) data for training.
  The resulting (a,b) are applied to 10 pairs of (x,y) test data.

  Training data shape:  (100, 2)
  Test data shape:  (10, 2)
First 3 training items: [0.598499 0.60241  0.528659]
First 3 training targets: [2.03206 2.44061 2.37146]

Training the model:

step 0: training cost = 2.94896
step 10: training cost = 2.39731
step 20: training cost = 2.00303
step 30: training cost = 1.71348
step 40: training cost = 1.49514
step 50: training cost = 1.32633
step 60: training cost = 1.19275
step 70: training cost = 1.08486
step 80: training cost = 0.996098
step 90: training cost = 0.921917
step 100: training cost = 0.859064

Testing the model:


1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 48ms/step - loss: 0.4289
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 51ms/step - loss: 0.4289

Results:

Test cost: 0.42890021204948425
Weights =  [[1.810066]]
Bias =  [1.3694174]

  Graphics saved as "montana.png"

montana():
  Normal end of execution.
Thu Mar 28 23:18:13 2024