Generative Deep Learning by David Foster

• A generative model must also be probabilistic rather than deterministic, because we want to be able to sample many different variations of the output, rather than get the same output every time.

• A generative model must include a random component that influences the individual samples generated by the model.

Representation Learning

• Suppose you wanted to describe your appearance to someone who was looking for you in a crowd of people and didn’t know what you looked like. You wouldn’t start by stating the color of pixel 1 of a photo of you, then pixel 2, then pixel 3, etc. Instead, you would make the reasonable assumption that the other person has a general idea of what an average human looks like, then amend this baseline with features that describe groups of pixels, such as I have very blond hair or I wear glasses. With no more than 10 or so of these statements, the person would be able to map the description back into pixels to generate an image of you in their head.

If you use word tokens:

• All text can be converted to lowercase, to ensure capitalized words at the start of sentences are tokenized the same way as the same words appearing in the middle of a sentence. In some cases, however, this may not be desirable; for example, some proper nouns, such as names or places, may benefit from remaining capitalized so that they are tokenized independently