Course 1 - Dr. Bah

Lectures

Lecture 1: ANN (slides)(videos)

• Linear Neural Networks – linear regression, logistic regression, …

• Multi-layer Perceptron (MLP) - hidden layers, activations, forward & backward propagation, …

• Issues around training – model selection, underfitting, overfitting, weight decay, dropout, …

Lecture 2: CNN (slides)(videos)

• From fully-connected layers to convolutions

• Convolutions & Cross-correlations, padding, stride, …

• Multiple input and multiple output channels

• Pooling

• Popular CNNs – LeNet, AlexNet, VGG, etc

Lecture 3: RNN (slides)(videos)

• Sequence modeling

• Recurrent Neural Networks

• Backpropagation through time

• Popular RNNs – GRU, LSTM, Deep RNNs, Bidirectional RNNs, Encoder-decoder architectures, …

Lecture 4: Autoencoders & GAN (slides)(videos)

• Dimensionality reduction & autoencoders

• Variational autoencoders

• Generative adversarial networks

• Deep Convolutional GANs

References

[0] Python – Keras & Tensorflow ready

[1] Dive into Deep Learning Book (https://www.amazon.science/deep-learning-book-adds-pytorch-tensorflow)

[2] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning, MIT Press (2016). Soft-copy (pdf) available at http://www.deeplearningbook.org.

[3] M. A. Nielsen, Neural Networks and Deep Learning, Determination Press (2015). Website accompanying the book: http://neuralnetworksanddeeplearning.com

[4] Online TensorFlow and Keras tutorials at https://www.tensorflow.org