Neural networks

Components of a neuron: weights, biases; trainable/learnable parameters

$x^{(t)}$ : 제곱 아님. t번째라는 뜻
$sign()$ : 0보다 크거나 같으면 +1 / 0보다 작으면 -1
$x_{n}^{(t)}$ : w에 대한 미분 ($\frac{d}{dw} (wx) = x$)
$y_{t}$ : 얼마만큼 조정?
$1[y^{(t)} \neq \hat{y}^{(t)}]$ : 조정 필요한지 여부 확인

Neural network

dimensions of input/output, weights, bias; trainable/learnable parameters

forward propagation

backward propagation (i.e. gradient propagation and parameter updates)

Activation functions: sigmoid, tanh, ReLU, softmax

Regression vs classification objective

L2 loss, L1 loss, cosine loss

Cross-entropy loss, one-hot encoding