Course Address: Https://class.coursera.org/ntumltwo-002/lecture
Important! Important! Important!
1. Shallow-layer neural networks and deep learning
2. The significance of deep learning, reduce the burden of each layer of network, simplifying complex features. Very effective for complex raw feature learning tasks, such as machine vision, voice.
In the following digital recognition, the pixel feature is converted to the stroke feature for learning without knowing the whole digital feature directly
3. Problems and key technologies for deep learning. With the increase of the number of layers of neural network, a variety of neural network structures can be designed, and it is difficult to choose a useful structure for the problem. The complexity of the model and the amount of computation also become very large.
Ms. Lin thinks regularization (regularization) and initialization (initialzation) are the key techniques for designing good deep learning.
4.
Machine learning techniques-deep learning