If I have a input picture of a number, what are weights and bias present on that picture?

[thinh-doan]

If I have a input picture of a number, what are weights and bias present on that picture? Thank you!

[Deetjepateeteke]

There are no weights or biases present on the input picture. The weights and biases are parameters of your neural network so it can learn patterns.

To view those parameters, you can do:

params = model.state_dict()
print(params)

The input picture is a 28x28 grid of decimal numbers ranging from 0 to 1. Those numbers represent the luminance of every pixel in the image:

• close to 0 -> black
• close to 1 -> white

These pixel values are NOT weights or biases.