Inside Neural Numbers
films
Licence
Crédits
Inside Neural Numbers combines short clips created for the MOOC “AI Explorables for Schools” on the KI Campus platform and the upcoming exhibit “AI Suitcase”.
Contents of the video:
0:00 - 0:27: In our number recognizer, the input for the neural network is a 28 x 28-pixel image. 28 times 28 pixels results in 784 individual pixels. The neural network assigns a number between 1 (for black) and 0 (for white) to each point. Decimal places are used for shades of gray.
0:27 - 1:04: A neural network is built up of layers. The input layer is built from the values of the input pixels.
1:05 - 1:39: There is a hidden layer and the output layer in our neural network. Our hidden layer has 100 neurons; the output layer has 10 (one for each number from 0 to 9). The flow of information has a clear direction, which is also known as feed-forward.
1:40 - 3:05: Neural network with example values.
Fichiers
Plus de films
