Neural networks
In this video I’m going to be talking - at a very high level - about how graphs can be used for working in different problem domains. Rather than focusing on formalism or definitions, I want to show you how graphs can be used. Hopefully this will help you use graphs to solve problems that you encounter.
In this video we’ll be seeing how graphs allow the complexity of Neural Networks.
Important: You will not be tested on the problem domain! I do not expect you to learn about neural networks. If this video is helpful to you, great! If not, don’t worry about it!
Eleven-minute video
You can also view this video on YouTube
Key Points
Neural networks
- A neural network is a network (graph) of nodes (“neurons”) that each perform a simple function.
- Each neuron takes the weighted sum of its inputs, and outputs a value
- By combining these together in complex ways, and with the correct weights, you can solve incredibly difficult problems
- Algorithms have been written that allow a neural network to learn it’s own weights - so they get better with training.
Weighted graphs
- In a weighted graph, edges are associated with a weight or a value.
- The weight might represent importance, or capacity, or strength, or some other property.
- Weighted graphs help with solving problems like finding the maximum flow through a network of pipes.
Questions
Check your Understanding
Which of the following problems could be represented using a weighted graph>?
Summary
In this section we saw how neural networks make use of graphs to combine simple functional components to solve difficult tasks.
In the final section we are going to see how trees can be used in game engines.