扩散模型的前向过程与反向过程分别做什么？
What do the forward and reverse processes of a diffusion model do?
前向过程逐步往真实图像里加入高斯噪声，直到变成纯噪声；反向过程训练一个神经网络从噪声中一步步预测并减去噪声，最终“雕”出清晰图像。采样就是反复运行反向过程。
The forward process gradually adds Gaussian noise to a real image until it is pure noise; the reverse process trains a network to predict and remove noise step by step, 'sculpting' a clean image. Sampling repeatedly runs the reverse process.
难度 专业 · 适合 大众
CC BY 4.0 · 来源待核验 · Source pending review
来源 · Source
出处备注：著录含书名，但缺版次 / 出版社 / 年份中的一部分，需编辑补全后才能做独立比对。
Ho、Jain、Abbeel,《Denoising Diffusion Probabilistic Models》,NeurIPS,2020
J. Ho, A. Jain, P. Abbeel, Denoising Diffusion Probabilistic Models, NeurIPS 2020
#扩散 #生成
为什么扩散模型生成图像往往比 GAN 更稳定？
Why is diffusion often more stable than GANs for image generation?
GAN 靠生成器与判别器对抗训练，容易模式崩溃且难调参；扩散模型把生成拆成许多小步去噪，训练目标简单（预测噪声），因此收敛更平稳、模式覆盖更全。
GANs pit a generator against a discriminator and easily collapse or need careful tuning; diffusion breaks generation into many small denoising steps with a simple noise-prediction loss, so training is steadier and covers modes better.
难度 进阶 · 适合 大众
CC BY 4.0 · 已核验 · Verified 2026-08-23 · 2 条独立证据
来源 · Source
Russell & Norvig,《Artificial Intelligence: A Modern Approach》第4版,Pearson,2020
S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Pearson, 2020
独立核验依据 · Independent verification: 核验依据 2 ARCHIVE 标识 isbn_9788177583670：Artificial Intelligence: A
#扩散 #GAN
文生图模型是怎么把一段文字变成一幅画的？
How does a text-to-image model turn words into a picture?
模型先把你的文字编码成含义向量，再以一张随机噪点图起步，一边理解文字含义一边逐步抹去噪点、显出轮廓与色彩，最终得到符合描述的画面。
The model encodes your text into a meaning vector, starts from random noise, and while 'reading' the text gradually removes noise to reveal shapes and colors, ending with a picture that matches the description.
难度 入门 · 适合 大众
CC BY 4.0 · 已核验 · Verified 2026-08-23 · 2 条独立证据
来源 · Source
Russell & Norvig,《Artificial Intelligence: A Modern Approach》第4版,Pearson,2020
S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Pearson, 2020
独立核验依据 · Independent verification: 核验依据 2 ARCHIVE 标识 isbn_9788177583670：Artificial Intelligence: A
#文生图 #扩散