What You Always Wanted to Know About Deep Learning, but Were Afraid to Ask

Explore the fundamentals of Deep Learning in this comprehensive 58-minute conference talk from NDC Conferences. Gain a clear understanding of key concepts such as back-propagation, gradient descent, loss functions, and optimizers. Learn how to train a deep learning model for object recognition through a practical example. Discover the inner workings of neural networks, including hidden layers, weights, biases, and activation functions. Understand the importance of evaluating performance using cross-entropy and the role of optimizers in back-propagation. Delve into the process of initializing weights, updating them using Stochastic Gradient Descent, and the significance of partial differentials. Get introduced to TensorFlow and Keras as tools for implementing deep learning models. Perfect for those looking to demystify AI and machine learning concepts, this talk provides a solid foundation for beginners and a refresher for experienced practitioners in the field of Deep Learning.

Intro
Agenda
AI, Machine Learning, and Deep Learning
What is Deep Learning?
Implementing Deep Learning using Neural Networks Outputs
Inputs and Outputs in a Neural Network
Hidden Layer(s)
Weights and Biases
Calculating the Result of a Node (Forward Propagation)
Feeding the Result of a Node to an Activation Function
Categories of Activation Functions
Binary Step Function
Analogy
Use of Sigmoid Activation
Non-Linear Activation
Evaluating Performance
Cross Entropy
In Summary Activation Function and Loss Function
Using an Optimizer
Back Propagation
A walkthrough
Initializing the Weights
Significance of the Partial Differentials
Updating the Weights using Stochastic Gradient Descent
In Summary Activation Function, Optimizer, and Loss Function
TensorFlow and Keras