For accuracy

We expected the output [1 0 0] but we got the output [0.67 0.53 0.52]. Now we want to improve the net, so that we really get the desired output [1 0 0] in the future. To achieve our goal, we will use the faulty output to modify the weights and biases of your net. 1. We subtract the expected output from the faulty one: [0.67 0.53 0.52] - [1 0 0] = [-0.33 0.53 0.52] This shows us that the first position was 0.33 to small und position 2 and 3 were 0.53 and 0.52 to big. 2. We use these values to modify the output of the previous layer: We simply Matrix multiply the calculated difference [-0.33 0.53 and 0.52] to the output of the hidden layer. This means, our hidden layer will now actually be even more incorrect. My matrix multiplying the [-0.33 0.53 0.52] to the output, we pushed the output even further into delivering us results that have this exact error. We basically told the net "hey your output was wrong by -0.33. Repeat the exact same calculation next time and then add another -0.33 on top of that!" We do not want that. 3. We multiply our matrix by -1 to Inverse each entry. Now we tell our net "Hey your output was wrong by -0.33. Make sure to add +0.33 next time!" Maximizing the error means that the original -0.33 would have increased the difference between our expected and actual output.

Well , your brain is basically a complex neural network Plus, our body isn't us; our brain is us. We are just a complex meat neural network controlling a big fleshy, meaty and boney body.

If anyone can tell me what neural networks on my phone is please let me know. Bye the way, I've been hearing voices that say they are going to kill me, they started at Granville medical centre, and pretty much seems like everyone working for Granville county is in the know and I'm not knowing sht