Machine and Deep Learning Using MATLAB

6 Pretrained Neural NetworksTransfer Learning

Deep Learning: Image Networks

Deep learning is a subset of machine learning (or artificial intelligence in a broader sense) that teaches computers to do what humans naturally do: Learn from experience. Image networks are intrinsically neural networks. Deep learning employs neural networks to directly learn useful representations of features from data. Neural networks are inspired by biological nervous systems and combine multiple nonlinear processing layers using simple elements operating in parallel. Deep learning models can achieve cutting-edge accuracy in object classification, sometimes outperforming humans.

Deep Learning Toolbox includes simple MATLAB® commands for building and connecting deep neural network layers. Even without prior knowledge of advanced computer vision algorithms or neural networks, examples and pretrained networks make it simple to use MATLAB for deep learning.

Transfer learning is frequently employed in deep learning applications, where a pretrained network serves as a foundation for acquiring knowledge in a new task. The process of fine-tuning the network through transfer learning proves to be considerably faster and simpler compared to training it from the beginning. By utilizing a smaller dataset of training images, we can promptly train the network to grasp a new task. The key benefit of transfer learning lies in the fact that the pretrained network has already acquired a comprehensive set of features ...