Section 2 : Deep learning

Commentary

Section Goals

• To introduce the basic concepts and methods of deep learning.
• To discuss several models, algorithms and applications of deep learning.
• To discuss the benefits and challenges of deep learning.

Learning Objectives

Learning Objective 1

• Outline the basic concepts and models of deep learning.
• Explain how deep learning can help learn representations from data.
• Describe a basic training algorithm and strategy such as SGD and layer-wise training.
• Discuss a few most recent AI breakthroughs achieved by machine learning and deep learning.
• Explain the following concepts or terms:
  • Deep learning
  • Representation learning
  • Auto-encoders
  • Deep belief networks (DBNs)
  • Deep Boltzmann machines (DBMs)
  • Convolutional neural networks (CNNs)
  • Recurrent neural networks (RNNs)
  • Stochastic gradient descent (SGD)
  • Deep reinforcement learning

Objective Readings

Required readings:

LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.

Arel, I., Rose, D. C., & Karnowski, T. P. (2010). Deep machine learning—a new frontier in artificial intelligence research. IEEE Computational Intelligence, 5(4), 13-18.

Bengio, Y., Courville, A., & Vincent, P. (2013). Representation learning: A review and new perspectives. IEEE Trans. on Pattern Analysis and Machine Intelligence, 35(8), 1798-1828.

Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., ... & Petersen, S. (2015). Human-level control through deep reinforcement learning. Nature, 518(7540), 529-533.

Supplemental Readings:

Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT press. Available at: https://www.deeplearningbook.org/

Bengio, Y. (2009). Learning deep architectures for AI. Foundations and trends® in Machine Learning, 2(1), 1-127.

Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., Van Den Driessche, G., ... & Dieleman, S. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587), 484-489.

Objective Questions

• What are the main differences between the deep learning and traditional neural networks, with respect to algorithms and performance?
• Why is learning representations from data a key to success for many AI and machine learning applications?
• What are the limitations of deep learning? How can other branches of AI help deal with them?

Objective Activities

• Explore the Internet to find some recent breakthroughs in AI applications achieved mostly by machine learning, especially deep learning, and discuss them in the online course conference.
• Search and explore the following deep learning tools.
  • PyTorch (https://pytorch.org/ or https://github.com/pytorch/pytorch)
  • Tensorflow (https://www.tensorflow.org/ or https://github.com/tensorflow/tensorflow)
  • MXNet (https://github.com/apache/incubator-mxnet)
  • Keras (https://keras.io/)
• Explore deep learning applications or prototype systems for further research.