提出可微向量量化新方法，实现端到端训练且无需额外损失。
DiVeQ: Differentiable Vector Quantization Using the Reparameterization Trick
- 将量化视为添加误差向量，前向保持硬分配，反向梯度可流动。
- 在图像压缩、生成与语音编码任务中，重建质量优于现有方法。
- 无需温度调度或辅助损失，适合需要高效训练的模型场景。
向量量化广泛用于深度模型，但其硬分配会阻断梯度，阻碍端到端训练。本文提出DiVeQ，将量化建模为添加一个模拟量化失真的误差向量，前向过程保持硬分配，同时允许梯度传播。我们还提出一种空间填充变体（SF-DiVeQ），将输入分配到由码本向量连线构成的曲线上，降低量化误差并实现码本的完全使用。两种方法均支持端到端训练，无需辅助损失或温度调度。在VQ-VAE图像压缩、VQGAN图像生成及DAC语音编码等任务中，基于多个数据集的实验表明，所提方法在重建质量和样本保真度上均优于其他量化方案。
原文摘要 · Abstract (English)
Vector quantization is common in deep models, yet its hard assignments block gradients and hinder end-to-end training. We propose DiVeQ, which treats quantization as adding an error vector that mimics the quantization distortion, keeping the forward pass hard while letting gradients flow. We also present a space-filling variant (SF-DiVeQ) that assigns input to a curve constructed by the lines connecting codewords, resulting in less quantization error and full codebook usage. Both methods train end-to-end without requiring auxiliary losses or temperature schedules. In VQ-VAE image compression, VQGAN image generation, and DAC speech coding tasks across various data sets, our proposed methods improve reconstruction and sample quality over alternative quantization approaches.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。