DEEP LEARNING: HANDS ON

A joint activity of
METU Electrical and Electronics Engineering
METU-HU Neuroscience and Neurotechnology Ph. D. Program
Intel Nervana AI Academy METU Branch
METU VISION Research Lab

Tutorials by:
Egemen SERT
Barkın TUNCER
Sefa Burak OKCU
Eren HALICI
Fourough GHARBALCHI
Meltem ATAY

Mentor:
Ugur HALICI, Prof. Dr.

A 12-weeks-long Deep Learning tutorial series including workshops for practice will be initiated on October 25, 2017 Wednesday (18:40-20:30 at METU U3 Hall).

Tentative program:

1. Introduction: What is Deep Learning? Why do we need data?
What can be done with DL? What is loss function intro

2. Neural Networks: What is regression, classification?
What is Neural Network, What does the Neural Network learn?

3. Neural Networks Training: What is Backpropagation? What is Gradient Descent?
How does the Neural Network learn?

4. WORKSHOP I: NN implementation, SGD, Dropout, Batch Normalization, Adam Optimizer, Train/Test Splits, Training CIFAR-10

5. Convolutional Neural Networks: What is Convolution? What is kernel/filter? What is pooling?
What is dilation? What is segmentation?

6. CNN Architectures: AlexNet, VGGNet, ResNet, GoogLeNet, Highway Net, UNet, DenseNet, R-CNN, Faster-R-CNN…

7. WORKSHOP II: segmentation example? face detection example? (with Keras Framework)

8. Recurrent Neural Networks: What is an RNN? Why do the gradients vanish? How stable is vanilla RNN? What is gating? What is LSTM & GRU?

9. Attention: What is Attention? What is External Memory? Neural Machine Translation, Image Captioning (w/ attn)…

10. WORKSHOP III: English-French translation, video search, image captioning (with Keras Framework)

11. Autoencoders: What is AE, GAN, VAE, Style Transfer

12. Deep Reinforcement Learning