46. Deep Learning vs Classical ML: When to Use Which
46.2 When Deep Learning Helps
Why. Some inputs are awkward for hand-made features: pixels, waveforms, long reviews.
What. Deep learning helps when patterns are hierarchical and local (edges→shapes→objects) or contextual in sequences. You need more data and more compute in return.
How.
DL-friendly signals
- Images / video frames → CNN path (43–44).
- Long text → transformers (45–46) / Gen AI for LLM apps.
- Huge clickstream with complex interactions — sometimes DL, sometimes boosted trees; test both.
print("Product photo classify -> CNN")
print("Chat-style assistant -> Gen AI course / LLM APIs")