Summary of Week 5 -- Capsule 2 -- Training Neural Networks

00:00:00 - 00:20:00

This video discusses the back propagation algorithm for training neural networks. It explains how to calculate the gradient of the objective function with respect to a particular parameter and use this to update the parameter. The video also discusses the gradient descent algorithm and how adding more hidden layers can improve the accuracy of a neural network.

• 00:00:00 In this video, the author introduces the back propagation method for learning the weights of a neural network. This method is based on the idea that we want to change the parameters in order to improve our performance.
• 00:05:00 The video discusses how to calculate a gradient with respect to a particular parameter using the back propagation algorithm. The first step is to calculate the gradient of the objective function with respect to that parameter at step t at the beginning of the iteration. Next, for each parameter, the gradient is calculated and used to update the parameter. If the parameter is not changing, then the gradient will be zero, and the parameter will not be updated. If the parameter is changing, then the gradient will be used to update the parameter.
• 00:10:00 This video explains how to train a neural network using gradient descent. The gradient descent algorithm minimizes the squared error of the network by updating the weights according to the gradient of the squared error.
• 00:15:00 This video discusses the gradient descent algorithm, which is used to calculate the best fit for a neural network. The gradient descent algorithm is more computationally expensive than a single data calculation, and can produce more noisy results. The video also discusses how adding more hidden layers can improve the accuracy of a neural network.
• 00:20:00 This video explains how to optimize neural networks. It states that this can be done, but it depends on the data set and the algorithm used.