Neural Networks: Training

Question 1

Part 1

Answer 1

Feeding data - The network starts by taking in input data through the input layer.

Question 2

Part 2

Answer 2

Input data flows from the input layer through the hidden layers and ultimately, the output layer, where the NN makes its first prediction.

Question 3

Part 3

Answer 3

Calculating error. After making a prediction, the NN will then compare its prediction to the correct answer (extracted from its training data).
The difference between the prediction and the correct answer is calculated and standardized using a loss function.

(loss function measures how wrong the network’s predictions are for a single epoch; the goal is to make this error as small as possible)

Question 4

Part 4

Answer 4

Learning from our mistakes or specifically, backpropagation. This is when the NN will reflect on its errors and use that reflection to figure out how to adjust its neurons’ calculations (by adjusting weights) to make better predictions next time.

Question 5

Part 5

Answer 5

Repeating the process. This whole process of inputting data, making a prediction, calculating errors, and adjusting weights using backpropagation is repeated with many epochs of the NN.

Question 6

Part 6

Answer 6

Evaluating performance. After several epochs, the NN’s performance is evaluated to see if it has improved and accurately predicting for the task it was designed for.