{"ID":2862903,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2509.26469","arxiv_id":"2509.26469","title":"DiVeQ: Differentiable Vector Quantization Using the Reparameterization Trick","abstract":"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.","short_abstract":"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-fillin...","url_abs":"https://arxiv.org/abs/2509.26469","url_pdf":"https://arxiv.org/pdf/2509.26469v4","authors":"[\"Mohammad Hassan Vali\",\"Tom Bäckström\",\"Arno Solin\"]","published":"2025-09-30T16:17:21Z","proceeding":"cs.LG","tasks":"[\"cs.LG\"]","methods":"[\"Generative Adversarial Network\",\"Variational Autoencoder\"]","has_code":false}