(b) Explain how the filters contribute to the function of the CNN.View keyboard shortcutsEditViewInsertFormatToolsTable12ptParagraph

Convolutional Neural Networks (CNNs) use filters during the convolutional layers for feature extraction. Here's how they contribute to the function of the CNN:

1. Feature Identification: Each filter is responsible for identifying a specific feature in the input. For example, one filter might be designed to identify edges, another to identify circles, and so on. These features could be anything from simple shapes to complex patterns depending on the depth of the layer in the network.

2. Convolution: The filters are moved across the input image (or the output from a previous layer) to create a feature map or convolved feature. This is done by multiplying the values in the filter with the original pixel values of the image, which results in a single number. The process is repeated for every location on the input.

3. Non-Linearity: After convolution, a non-linear transformation (like ReLU) is applied to the feature maps. This is done to introduce non-linearity into the model, as the real-world data is non-linear and the aim of the CNN is to learn and represent this data.

4. Pooling/Subsampling: The feature maps are then passed through a pooling layer, where the CNN downsamples the feature maps to reduce their dimensionality. This helps to decrease the computational complexity of the network and to avoid overfitting.

5. Learning: During the training process, the CNN learns the values of these filters on its own, although we need to specify parameters such as the number of filters, filter size, architecture of the network, etc. The more diverse the images, the better the network can learn and generalize.

6. Classification: Finally, the output from the convolutional and pooling layers represent high-level features of the input. These output vectors are flattened and fed into a fully connected layer to classify the images.

In summary, filters in CNNs help in extracting different features from the input, which are then used to classify the input into various categories. They are the primary components that enable CNNs to automatically and adaptively learn spatial hierarchies of features.