What does “Deep Learning” mean?
A subset of machine learning that uses many-layer neural networks for representation learning.
How Deep Learning Works
Raw input: pixels or text
Layer 1: simple patterns
Layer 2: combined patterns
Layer 3+: abstract concepts
Prediction.
Practical Example
A deep network that recognizes cats learns edges in its first layers, then fur and ear shapes, and finally whole faces in later layers. No one programs those steps; the layers form them during training on many labeled photos.
Why It Matters
Deep learning drives most current progress in speech recognition, image analysis, and language models. It needs a lot of data and compute, which shapes who can build such systems.
Common Questions About Deep Learning
Q: What makes learning 'deep'?
The network has many stacked layers, each building on the output of the one before.
Q: Is deep learning the same as AI?
No. It is a subset of machine learning, which is itself a part of AI.
Q: When is deep learning not the best choice?
On small tabular datasets, simpler methods such as gradient-boosted trees often match or beat it and are cheaper to run.