The Big Picture Series: Concepts & Terminologies To Begin Your DL Journey With

Heard of Deep Learning but want your big picture digest? Kickstart your intuition with these concepts.

Knowledge representation often clouds the intuition if not understood well. Questioning nomenclature often is a sensible approach to building solid understanding.

Legend

  • Concept terms : big picture words, try to employ your imagination and trust your intuition when you read it. Resist thinking that you won't fully understand the concept until you go through all of its branches (the concept is the tree, just try to understand its general structure and why was it seeded there to begin with)
  • Technical terms : here we dive to the lower levels. Try to keep your peripherals on the big pictures when you digest these
  • Etymology : word origins.

[1] Deep Learning

Deep: Because the networks being constructed are of many layers (a simple network would comprise of 1 or 2 layers). Learning: The networks learn the features that make the thing being studied. (e.g. learning the edge shape of a dog's ear, the roundess of its nostrils, the color scheme of its skin, etc.).

[2] Neural Networks

Neural: Most of artificial intelligence is inspired from the biological aspects that make up human intelligence, the first attempt at these networks was called a Perceptron, its simple structure imitates the information flow in neurons (reprented as units within layers).

Networks: A network is a list of interconnected elements of any sort.

[3] Convolutional Neural Networks (CNNs)

Convolutional: from con- ‘together’ + volvere ‘roll’ yep, the input is being rolled together with a filter.

[4] Generative Adversarial Networks (GANs)

The idea is to have one network “generate” a lot of adversarial samples to another network for it to train upon.

[5] Sequence Models / Recurrent Neural Networks (RNNs)

[6] Reinforcement Learning