Cyan's Blog

Welcome!🎉 I’m Cyan Fu, a master student at Carnegie Mellon University studying Artificial Intelligence.

I write about anything and everything here, and my main interest lies at the intersection of AI and 3D vision. I hope you find something interesting here!

Check out the network at the bottom of the page!

Site under construction

I’m trying to build a new site with better LaTeX support and automatic Chinese <-> English translation with GPT4.

Sorry if some link is broken!

• 一个抽象而有趣的引理： Johnson Lindenstrauss Lemma
• 从二项分布到泊松分布再到指数分布-From Binomial Distribution to Poisson Distribution to Exponential Distribution
• 为什么方差的分母常常是n-1
• 酉矩阵为什么叫酉矩阵
• 如何对矩阵求导？
  • 矩阵的求导
  • D2L-4-矩阵求导
• Diffie-Hellman问题
• KL_Divergence-KL散度
• Gilbert Strang 深入浅出机器学习的矩阵知识
• Understanding Bayes’ Theorem

• What on earth is Logit
• 为什么Softmax回归不用MSE
• Kernel Regression
• Norm in Regularization - Intuition
• Fisher Linear Discriminant(Pattern_Classification-Chapter_4)
• Cross_Entropy-交叉熵

• Convolution-卷积，数学与ML含义
• 1x1卷积层有什么用
• 智能版加权平均：Attention
  • Kernel Regression and Attention
  • Attention Scoring Function
  • 图解Additive Attention
  • Scaled Dot-Product Attention
  • Seq2Seq with Attention - Bahdanau Attention
  • 图解Multi-Head_Attention
  • 图解Self-Attention
• 智能版加权平均 is All You Need: 图解Transformer
• 端到端学习-End_to_End_Learning-E2E
• 好的预测模型的特质
• What is a tensor
• 自动求导
  • 为什么反向传播比前向传播更高效
• MLP-多层感知机
• 权重衰减
• 数值稳定性
  • D2L-25-让训练更加稳定-Xavier初始化
• 经典模型

有效地背诵GRE单词

如何努力工作_Paul_Graham

• Feel free to explore the graph below!