Coursework publication notice / 课程作业展示说明
To respect course academic-integrity requirements, detailed numerical result tables and written summaries, assignment solution code, mathematical derivations, and original written answers are kept private. This public portfolio presents visual outputs, project context, and skills only.
为遵循课程对作业解答的发布限制，具体数值的表格与文字汇总、作业解答代码、推导及原始书面答案均保留在私有仓库。公开部分仅展示图表、项目概述与技能。
Original axes and annotations within the figures are retained. 图中的原有坐标刻度与数值标注保留。
Qingyang Ling · Stanford Online XCS236
A visual overview of my coursework in probabilistic modeling and deep learning, spanning autoregressive models, latent-variable generation, normalizing flows, adversarial learning, and score-based diffusion.
| Variational autoencoder | Gaussian-mixture variational autoencoder |
|---|---|
 |
 |
| GAN | Conditional GAN | WGAN-GP |
|---|---|---|
 |
 |
 |
| Diffusion generation | Diffusion generation |
|---|---|
 |
 |
| Topic | Visual evidence |
|---|---|
| Autoregressive models | Saved likelihood-distribution plots |
| Latent-variable models | Generated handwritten digits |
| Normalizing flows | Sample progression and saved loss curves |
| Adversarial learning | Clothing generation across saved training snapshots |
| Score and diffusion models | Generated faces and a saved resource-comparison plot |
Python, PyTorch, probabilistic reasoning, experimental evaluation, and documenting model assumptions. The saved figures show historical coursework outputs; no training or inference was rerun while preparing this gallery. The face examples use a supplied pretrained model and demonstrate sampling, rather than a face model trained by me.
This public repository contains images and a broad project overview. Completed code, mathematical derivations, written submissions, numerical result tables, data, and checkpoints remain in a separate private coursework archive. Original labels and numerical annotations inside the figures are retained; the figures are not redacted.
Display scope and provenance · Official course information
中文简介：课程学习与可视化成果展示。代码、推导、原始提交报告和数值表保留在私有仓库；这里展示原始结果图及项目概述。