lab02 Linear regression

Hypothesis

y = Wx + b

W (weight), b (bias)

import torch

x_train = torch.FloatTensor([[1], [2], [3]])
y_train = torch.FloatTensor([[2], [4], [6]])

W = torch.zeros(1, requires_grad=True) # required_grad -> 학습할 것이라고 명시
b = torch.zeros(1, requires_grad=True)
hypothesis = x_train * W + b

Compute loss

Mean Squared Error (MSE)

loss (Model이 정답과 얼마나 가까운가) 를 계산 -> Prediction (예측값) 과 Target(학습값) 의 차이를 제곱하여 평균

cost = torch.mean((hypothesis - y_train) ** 2)

Gradient descent

앞서 계산한 loss를 바탕으로 model 개선

optimizer = optim.SGD([W, b], lr=0.01)

optimizer.zero_grad() # gradient 초기화
cost.backward() # gradient 계산
optimizer.step() # gradient 개선

Full Training code

# Data 정의, Hypothesis 초기화, Optimizer 정의 (최초 1회)
x_train = torch.FloatTensor([[1], [2], [3]])
y_train = torch.FloatTensor([[2], [4], [6]])

W = torch.zeros(1, requires_grad=True)
b = torch.zeros(1, requires_grad=True)

optimizer = torch.optim.SGD([W, b], lr=0.01)

# Hypothesis 예측, Cost계산, Optimizer로 학습 (반복)
nb_epochs = 1000
for epochs in range(1, nb_epochs + 1):
    hypothesis = x_train * W + b
    cost = torch.mean((hypothesis - y_train) **2 )
    
    optimizer.zero_grad()
    cost.backward()
    optimizer.step()