Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Gentle introduction to deep learning

Gentle introduction to deep learning

A brief introduction to what deep learning is about, applications of deep learning and learning methodologies.

Rajika Imal

October 26, 2018
Tweet

More Decks by Rajika Imal

Other Decks in Programming

Transcript

  1. Artificial intelligence The theory and development of computer systems able

    to perform tasks requiring human level intelligence such as visual perception, speech recognition. - wikipedia
  2. Machine learning Sub-field in artificial intelligence that uses statistical techniques

    to give computer systems the ability to learn. - wikipedia - machine learning mastery
  3. Learning with labelled training data Example: • Classifying email as

    spam or ham. • Predicting the breed of a dog. Supervised learning - stats and bots
  4. Unsupervised learning Self discovering patterns in unlabelled data Example: •

    Clustering news articles in Google news. - oracle blogs
  5. Reinforcement learning Learning based on the reward given for actions

    performed. Example: • An agent playing atari. • Chess playing agent. - hackernoon
  6. Deep learning Subfield in machine learning. A technique inspired by

    how the human brain works. Term deep learning was introduced in 1986 by Rina Dechter.
  7. Fast forward to late 2000 • Big data • Huge

    compute power ◦ Advanced CPUs ◦ GPUs
  8. Applications of deep learning • Autonomous vehicles. eg: Tesla autopilot

    • Translation applications • Image tagging • Text generation • Colorization of black and white images • Music generation Tesla autopilot: https://bit.ly/2fTmW13 Fifa playing NN: https://bit.ly/2O8vIIz Video surveillance: https://bit.ly/2meiNaR
  9. Types of neural networks • Convolutional neural networks (CNN) •

    Recurrent neural networks (RNN) • Generative adversarial networks (GAN) • LSTM
  10. Top-down approach Starting from a framework and working towards creating

    models to be applied for sample scenarios. Example: • Spam detector • Image classifier to classify dogs and cats • Natural language generator for english Frameworks: • PyTorch, Tensorflow, Keras, Fast.ai
  11. Mixed-approach Combining top-down and bottom-up approaches to gain knowledge about

    theoretical concepts as well as the current frameworks used by developers and researchers.
  12. Who to follow? • Andrew Ng (Co-founder of Google brain

    / Coursera / Baidu AI) • Jeff Dean (Co-founder of Google brain) • Siraj Raval (School of AI / Content creator) • Andrew Trask (OpenMind) • Karpathy (Tesla) • Ian Goodfellow (Google brain) • Jeremy Howard (fast.ai)