Page 171 Boltzmann Scaling

The equation given should read:

Where the brackets represent the average at the current generation of the contained term.

Also the code listing is incorrect. You can download a new project file with the correct source here.

Page 199 Normalizing vectors

I must have been drinking too much coffee when I wrote this. I divided by the square of the length instead of the length! The text should read:

3/5 = 0.6 and 4/5 = 0.8

Therefore the normalized vector, N, is (0.6, 0.8)

Page 300.

I managed to get the signs of a couple of terms muddled up. The text should read (changes in bold):

Step three. Calculate the error value for the neuron in the hidden layer using the equation:

error = 0.5 x (1 – 0.5) x -0.125 x -0.00625 = 0.000195

Notice how I’ve used the updated weight computed in Step two for w_jk to calculate the error. This is not essential. It’s perfectly reasonable to step through the network calculating all the errors first and then adjust the weights. Either way works. I do it this way because it feels more intuitive when I write the code.

Step four. Adjust the weights going into the hidden layer. Again, going from left to right and using the equation:

New weight (bias) = 0 + 0.1 x 0.000195 x -1 = -0.0000195

New weight1 = 0 + 0.1 x 0.000195 x 1 = 0.0000195

New weight2 = 0 + 0.1 x 0.000195 x 1 = 0.0000195

Please contact me if you have found an error

bugs_at_aijunkie.com

Home