What are weighted connections?

What are weighted connections?

A network with weighted connections is a network where the ties among nodes have weights assigned to them.

A practical, real-life example is when you are deciding whether to buy ice cream or not. Let’s assume that the only factors that matter to you when buying ice cream are the following inputs:

  1. Strawberry flavor being available
  2. Being summer
  3. Having cash

But when strawberry flavored ice cream is out of stock you still buy it because you really love ice cream, but you almost never buy when it’s not summer. You can use your credit card so not having cash is often ok. In this case, you apply stronger weights to “being summer” when building your network.

  1. Strawberry flavor being available – Weight 1
  2. Being summer  – Weight 2
  3. Having cash – Weight 1

I hate having to go into maths, but I will keep it simple, I swear!

We have 4 weights (1+2+1), if we turn it into a percentage, each of this is the equivalent to 25% in your decision making when it comes to ice cream. So:

  1. Strawberry flavor being available – 25%
  2. Being summer  – 50%
  3. Having cash – 25%

Let’s say you are on a diet and you are only purchasing ice creams when at 75% of your criteria is met. It’s pretty safe to assume that you will only ever consume ice cream during summer time because you can never reach 75% with just the other criteria.

During summer you will only consume ice cream if you can pay cash or strawberry flavor is available. No chocolate flavor on credit for you!

This is a very, very simple example. It’s not uncommon for neural networks to have thousands of inputs.


What is next?