Notes for Prof. Hung-Yi Lee's ML Lecture 11: Why Deep?
Having the same total number of parameters, the deep, i.e. thin and tail, neural networks, performs better than the fat abd short networks. The reason is that deep neural networks benefits from modularization, so they use the data and the parameters more efficiecntly.
Notes for Prof. Hung-Yi Lee's ML Lecture 10: CNN
CNN: Convolutional Neural Networks
Notes for Prof. Hung-Yi Lee's ML Lecture 09: Tips for Deep Learning
Steps Toward a Successful DL Training
Notes for Prof. Hung-Yi Lee's ML Lecture 7: Backpropagation
For applying gradient descent on neural networks, there may be millions or more parameters, and we use backpropagation to compute the gradients efficiently.
Notes for Prof. Hung-Yi Lee's ML Lecture 6: Brief Introduction to Deep Learning
History of and Remarks on Deep Learning
Notes for Prof. Hung-Yi Lee's ML Lecture 5, Classification: Logistic Regression.
Logistic Regression
Notes for Prof. Hung-Yi Lee's ML Lecture 4, Classification: Probabilistic Generative Model.
What are Classification Problems
Notes for Prof. Hung-Yi Lee's ML Lecture 3, gradient descent.
Vanilla Gradient Descent Formulation
Notes for Prof. Hung-Yi Lee's ML Lecture 2, Sources of error.
Sources of error Bias Variance
Notes for Prof. Hung-Yi Lee's ML Lecture 1, Regression
Notations