Deep learning fundamentals workshop
1 like161 views
The document outlines the fundamentals of deep learning, discussing various neural network types such as feed forward, recurrent, and convolutional networks. It highlights practical applications in diverse industries, including self-driving cars and chatbots, and emphasizes best practices for building AI systems. Additionally, the document includes hands-on sessions for practical learning using datasets like iris and mnist.
Deep learning fundamentals workshop
- 1. Deep Learning Fundamentals Satnam Singh, PhD Chief Data Scientist May 29, 2018
- 2. Outline • What is deep learning? • Feed Forward Neural Network (FFNN) • Recurrent Neural Network (RNN)& LSTM • Convolutional Neural Network (CNN) • How to build a deep learning network?
- 3. Applications of Deep Learning • Self-Driving Cars • Alexa • Chatbots • Almost every industry has use cases • Pls, share use cases from your industry….
- 4. ML vs. DL
- 5. ML vs. DL
- 6. What is Deep Learning?
- 10. Single Neuron - “AND” node
- 11. Single Neuron - “OR” node How to compute the weights?
- 12. Forward Propagation - 1
- 13. Forward Propagation - 2
- 14. Back Propagation Source: Python Machine Learning Book by Sebastian Raschka
- 20. 1. Hands-on Session: Iris Dataset
- 21. 2. Hands-on Session: Mnist dataset
- 22. Feed Forward Neural Network Inputs and outputs are independent
- 24. Time Series/ Sequence Prediction Stock Market Prediction Text Sequence Prediction The colour of the sky _____ Speech Translation
- 25. Recurrent Neural Network Output is dependent on Previous output
- 26. Problems in RNN The colour of the sky _____ RNN can predict: BLUE I grew up in India… I speak fluent ____ RNN CANNOT predict HINDI
- 27. RNN - Memory Source: Colah’s blog
- 28. RNN - Missing Long Term Memory RNN has Vanishing Gradient Problem
- 29. Long Short Term Memory (LSTM) Networks
- 30. Information Flow in RNN Source: Deep learning with Keras - Chollet
- 31. LSTM - Carry Bits
- 32. LSTM
- 33. 3. Hands-on Session: CnC Detection
- 34. Command and Control Detection C&C domain examples: • DGA based: gvludcvhcrjwmgq.in, uqvwxfrhhwreddf.yt • non DGA based: thisisyourchangeqq.com, homejobsinstitute.biz Ransomware Malware Enterprise Network Main DB Webserver C&C server Data Command Attacker
- 35. Update Models DNS data Ranking of Malicious C&C domains C&C domains Classify benign vs C&C domains using LSTM C&C Detection: Pipeline
- 37. CNN: Data Flow
- 38. CNN
- 39. Convolution
- 40. Convolution
- 43. Dropout
- 44. Softmax
- 45. Back Propagation
- 46. Hands-on Session: Cat vs. Dogs
- 47. Best Practices for Building AI Systems
- 48. 1. Start Small and Simple - First cut of AI system need not be the best
- 49. 2. Capturing the domain knowledge, data pre- processing is important.
- 50. 3. Need to follow Software Engg practices and tools - logging, data flow diagrams, abstraction, design patterns
- 51. 4. Identify and define performance metrics
- 52. 5. Integration with UI and Data Infra layers needs to be ironed out, use data interfaces, AI layer should be isolated
- 53. @satnam74s Where to go from here: • “Awesome Deep Learning - Github” https://github.com/ChristosChristofidis/awesome- deep-learning https://www.linkedin.com/in/satnamsinghdatascientist/