Main Content

An introduction to this website and a brief personal profile.

A planning note on how Harmonic Convergence will evolve into Convergence, a standalone iPad learning app.

A worked Bayesian linear regression example showing the design matrix, posterior derivation, MLE vs MAP, and predictive uncertainty.

A worked 2-2-1 neural network example showing initialization, forward pass, backpropagation, and one gradient descent update.

I’ll just put them here first, will come back to make them tidy later

In this article, I’ll be introducing a detailed explanation to view Variational Inference & ELBO statistically and mathematically.

An article that shares my current toolbox for learning, reasoning, coding and so on. Audio Engineering will be another topic.

在二维合成线性可分数据上，加入少量标签噪声后，带固定学习率的 perceptron 是否仍然能明显优于常数分类 baseline？ Hypothesis: 即使存在少量标签翻转，简单 perceptron 仍能在二维合成数据上取得明显高于 majority-class baseline 的 balanced accuracy；但其稳定性会随噪声上升而下降。

“none”

This guide reproduces the working setup of transforming a markdown file into voice.