elbo

Diffusion Classifier leverages pretrained diffusion models to perform zero-shot classification without additional training

A PyTorch Implementation of Generating Sentences from a Continuous Space by Bowman et al. 2015.

ICLR25 | Official code base for Heavy-Tailed Diffusion with Denoising Levy Probabilistic Models (DLPM)

(TIP 2026, CCF-A) ELBO-T2IAlign: A Generic ELBO-Based Method for Calibrating Pixel-level Text-Image Alignment in Diffusion Models

Code for Adversarial Approximate Inference for Speech to Laryngograph Conversion

Coordinate ascent mean-field variational inference (CAVI) using the evidence lower bound (ELBO) to iteratively perform the optimal variational factor distribution parameter updates for clustering.

Variational Autoencoder (VAE), Evidence Lower Bound (ELBO) Implementation (PyTorch)

Differentiable Probabilistic Models

rewriting ELBO: variational inference using symbolic rewriting techniques

Computational Statistics Project (MVA)

A from-scratch implementation of a Variational Autoencoder for FashionMNIST image generation in PyTorch. / FashionMNIST 28×28 이미지 생성을 학습하는, PyTorch로 직접 구현한 Variational Autoencoder.

Latent-Augmented Discrete Diffusion Models

Implementation of a Denoising Diffusion Probabilistic Model with some mathematical background.

Replication of the research paper titled Auto-Encoding Variational Bayes.

This course is part of the USC Graduate Biostatistics Program and is designed for second-year and beyond students interested in designing and implementing computational inferential tools for research.

手写变分推断 (Mean-Field CAVI)：贝叶斯高斯混合 + 坐标上升变分推断，ELBO 闭合形式(含 digamma/lgamma)。零依赖单文件 HTML，18/18 无头不变量 + 17/17 UI 全绿；K=1 时 ELBO≡精确证据(Δ3.6e-15)，对称问题下界间隙≡ln2。

Machine-checked Lean 4 proof of the finite evidence lower bound (ELBO), with learning resources

Factorized variational approximation using a univariate Gaussian distribution over a single variable x.

A pragmatic exploration of Generative AI comparing traditional Autoencoders with Variational Autoencoders (VAEs) on the MNIST dataset, contrasting standard amortized inference with Bayesian MAP latent optimization.

Unsupervised Learning