M6-049M6: Multimodal & Generative ModelsDiffusion Models Foundations (DDPM)Hard
⚡Flashcard Drill➔Mastery:
Velocity v-Prediction vs Epsilon-Prediction in Diffusion Models: SNR Stability and Extreme Noise Regimes
🎯 Active Recall Self-Check (Draft First, Verify After)
Login-free practice tracker: recall core answer, audit pitfalls, and log review status.
Step 1 · Draft your concise answer & assumptions
Step 2 · Check Pitfalls & Follow-up Inquiries
📐 Mathematical Definition
⚡ One-Sentence Core Answer
Velocity v-prediction parameterizes diffusion networks to predict a balanced linear combination of clean signal and noise, preventing gradient explosion and numerical collapse in high-resolution, zero-terminal SNR regimes.
📌 Key Takeaways
- •Epsilon-prediction breakdown: predicting noise becomes ill-conditioned in extreme noise regimes (), where small errors in cause explosive swings in estimated
- •V-prediction formulation: parameterizes network to predict velocity , smoothly interpolating between noise at and data signal at
- •Zero terminal SNR stability: v-prediction is mandatory for training models with zero terminal SNR, enabling high-contrast generation and arbitrary resolution scaling
📐 Mathematical Derivations
Mathematical Mechanism: 1. The Three Target Parameterizations: For latent state , with (where and ): (a) $\epsilon$-prediction: Network outputs . Estimated data state is: (b) $x_0$-prediction: Network outputs . Estimated noise state is: (c) $v$-prediction (Salimans & Ho, 2022): Network outputs velocity , defined as the angular velocity along the rotation trajectory: 2. Inversion Under v-Prediction: Given predicted velocity , both and are recovered symmetrically without dividing by near-zero coefficients: Because and , the transformations are orthonormal rotations, guaranteeing bounded gradients across all . 3. The Failure of $\epsilon$-Prediction at Zero Terminal SNR: At , zero terminal SNR sets . Under -prediction: Under -prediction: The formulation is perfectly stable and well-conditioned.
🏭 Production Trade-offs
Deep Dive & Engineering Trade-offs: ① Why Modern Diffusion Architectures Adopt v-Prediction: Early diffusion models (SD 1.5, standard DDPM) used -prediction because they avoided with non-zero terminal SNR schedules (e.g., ). However, non-zero terminal SNR prevents the model from generating true pure black or pure white images. Modern foundation models (SDXL 2.1, Stable Diffusion 3, Flux) enforce zero terminal SNR and require -prediction or Flow Matching to maintain training stability. ② Distillation Compatibility: Progressive distillation and consistency distillation collapse when applied to -prediction models due to accumulated step errors at high noise. Distillation algorithms converge with higher stability when operating on -prediction checkpoints. ③ Flow Matching Equivalence: In continuous Flow Matching with linear interpolation , the target vector field is . Velocity -prediction is the trigonometric spherical equivalent of linear flow matching velocity fields. ④ Interview Strategy: Formulate the three parameterizations (, , ), write down the velocity definition , show how and are recovered via orthonormal rotation without division, and explain why -prediction breaks down at zero terminal SNR.
⚠️ Common Interview Pitfalls
- ✕Attempting to implement zero terminal SNR noise schedules on -prediction models, causing training divergence due to division by zero
- ✕Swapping a pre-trained -prediction model into a -prediction inference pipeline without mathematical conversion
- ✕Assuming -prediction changes the model backbone architecture; it modifies only the target regression variable and loss formulation
🎯 Interviewer Follow-up Inquiries
- ?Why does -prediction become mathematically ill-conditioned when enforcing zero terminal SNR ()?
- ?How does velocity -prediction establish a mathematical bridge between discrete DDPM and continuous Flow Matching?
⚡
Test muscle memory on this topic via 3-Layer Flashcards
Active Recall: Mentally formulate Prompt ➔ Math Derivation ➔ Benchmark Implementation in 3 minutes.
📚