The limitations of deep learning

[manuzhang]

Article

• https://blog.keras.io/the-limitations-of-deep-learning.html

Notes

• In general, anything that requires reasoning—like programming, or applying the scientific method—long-term planning, and algorithmic-like data manipulation, is out of reach for deep learning models, no matter how much data you throw at them

  This is because a deep learning model is "just" a chain of simple, continuous geometric transformations mapping one vector space into another.

• "Adversarial examples" could easily trick deep learning models to misclassification since they don't have any understanding of input as humans do

• Human can do "extreme generalization"

  an ability to adapt to novel, never experienced before situations, using very little data or even no new data at all

  while deep nets can only map input to output ("local generalization"), which fails if inputs differ slightly from training time.

[manuzhang]

The author said,

This post is targeted at people who already have significant experience with deep learning (e.g. people who have read chapters 1 through 8 of the book). We assume a lot of pre-existing knowledge.

so I will revisit this article when I acquire more knowledge.