DDPM Score Matching and Distribution Learning