date | title | id |
---|---|---|
2021-04-26 11:54:00 -0700 |
Why are ANNs arranged in Layers? |
2021-04-26t18-54-00z |
Apart from simplifying how we reason and think about them, ANNs are typically arranged in layers to aid the underlying algorithms to reason about the problem in terms of sub-problems.
The hope is that by breaking down the network into sections (layers), each section or subgroup of sections settles into tackling a particular sub-problem whose solution can help the entire network in finding the right global solution.
For example, in image recognition tasks, it is often found that individual convolutional layers approximate certain functions such as the sobel operator for example.