COVER STORY u ANALOG DEVICES
pattern recognition process. A decision threshold is used for this. Another difference is that a pattern
recognition machine is not equipped with fixed rules. Instead, it is trained. In this learning process, a neural network is shown a large number of cat images. In the end, this network is capable of independently recognising whether there is a cat in an image or not. The crucial point is that future recognition is not restricted to already known training images. This neural network needs to be mapped into an MCU.
WHAT EXACTLY DOES A PATTERN RECOGNITION MACHINE LOOK LIKE ON THE INSIDE? A network of neurons in AI resembles its
As mentioned, CNNs are used for pattern
recognition and classification of objects contained in input data. CNNs are divided into various sections: one input layer, several hidden layers, and one output layer. A small network with three inputs, one hidden layer with five neurons, and one output layer with four outputs can be seen in Figure 2. All neuron outputs are connected to all inputs in the next layer. The network shown in Figure 2 is not able to
process meaningful tasks and is used here for demonstration purposes only. Even in this small network, there are 32 biases and 32 weights in the equation used to describe the network. A CIFAR neural network is a type of CNN that is widely used in image recognition tasks. It consists of two main types of layers:
The main difference between convolutional neural networks and other types of networks is the way in which they process data. Through filtering, the input data are successively examined for their properties. As the number of convolutional layers connected in series increases, so does the level of detail that can be recognised.
Figure 1: A neuron with three inputs and one output
biological counterpart in the human brain. A neuron has several inputs and just one output. Basically, such a neuron is nothing other than a linear transformation of the inputs - multiplication of the inputs by numbers (weights, w) and addition of a constant (bias, b) - followed by a fixed nonlinear function that is also known as an activation function. This activation function, as the only nonlinear component of the network, serves to define the value range in which an artificial neuron fires. The function of a neuron can be described mathematically as
where f = the activation function, w = weight, x = input data, and b = bias. The data can occur as individual scalars, vectors, or in matrix form. Figure 1 shows a neuron with three inputs and a ReLU activation function. Neurons in a network are always arranged in layers.
www.irish-manufacturing.com
Figure 2: A small neural network Irish Manufacturing October 2023 11
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50
