∑∫π∞√Δλμσθφ∂∇≈≠≤≥αβγωΩΣ∏⊕⊗ℝℕℤ∑∫π∞√Δλμσdef train(model, X, y):import numpy as npfrom sklearn import svmtensor.shape == (n, d)gradient = ∂L/∂wP(A|B) = P(B|A)P(A)/P(B)for epoch in range(N):loss.backward()SELECT * FROM data;σ(z) = 1 / (1 + e^-z)y = Wx + bargmin ||Ax - b||²class Net(nn.Module):df.groupby("id").mean()