So glad you are doing these! A lot of material is out there, but yours is encyclopedic-level comprehensive, fast, and easy to digest. Please keep those videos coming! . I'm still just a little confused about how Self-Learning takes unlabeled data (as with missing words in sentences) and creates labels for them. Does a human select a word from the list of AI-proffered suggestions, and that then becomes the label for the data-point (the sentence)?
@IntuitiveML
2 жыл бұрын
Creating labels for missing words is straightforward: You start with a database of complete sentences, for example, all text in Wikipedia. Then, for each sentence, you remove a word (or more), and save it. This word is your label, and the sentence is everything else. So given the sentence "I like watching IntuitiveML", you would (randomly) choose a word to remove, (for example: like). Then, your sentence with a missing word is "I __ watching IntuitiveML", and your label for the blank space is "like". You removed the word originally, so you know what it is.
Пікірлер: 6