Basics Flashcards
(8 cards)
1
Q
What are some classification usecases in NLP?
A
- Sentiment
- Part-of-speech tagging: noun, verbs…
- Named entity: find names of persons
2
Q
What is the one-hot vector?
A
- Vector of size N (size of vocab). N-1 are 0s and one 1 vWalue on the ith position.
No ordering, no context (mouse vs mouse)
3
Q
What are some tasks on graphs?
A
- Node classification
- Link prediction
4
Q
Describe a NN with one hidden layer
A
- Linear classifier with activation function at the end (sigmoid, tanh, RelU..)
5
Q
Function of tanh and how does it look like
A
𝑓 𝑥 = tanh 𝑥
6
Q
Function of sigmoid and how does it look like
A
7
Q
Function of ReLU and how does it look like. When is it used?
A
Soft and hard
8
Q
Explain Multi-layer NN. How is it non-linear? What are its functions?
A
Because of the non-linear activation functions
