He-Zhe (Mac) Lin 林鶴哲

I'm a junior software engineer at Qualcomm in Taiwan.

Before joining Qualcomm, I did my master at the Department of Computer Science and Information Engineering, National Taiwan University, advised by Professor Chih-Jen Lin. During the time, my research focused on developing algorithms for large-scale machine learning problem, particularly on extreme multi-label classification.

CV  /  Linkedin  

profile photo

Research
Illustration of k-means clustering: points grouped into three clusters by feature 1 and feature 2.
SparseKmeans: High-Performance Clustering on Sparse Data
Khoi Nguyen Pham Dang*, He-Zhe Lin*, and Chih-Jen Lin
CIKM, 2025
project page

Developed a high-performance K-means implementation using GraphBLAS for sparse matrix operations, achieving up to 10x speedup over scikit-learn on sparse data.

On the Weight Density of L2-Regularized Linear Classification and Regression
He-Zhe Lin, Zhi-Bao Lu, Sheng-Wei Chen, Cheng-Hung Liu, and Chih-Jen Lin
AISTATS, 2026
Slide illustrating inherent pruning in a tree-based linear model: features with no values in a data subset can be dropped from that node's weight vector.
Exploring Space Efficiency in a Tree-Based Linear Model for Extreme Multi-Label Classification
He-Zhe Lin, Cheng-Hung Liu, and Chih-Jen Lin
EMNLP, 2024
Computational graph for automatic differentiation: x1, x2 feed into v1..v5 to compute f(x1,x2).
A Step-by-Step Introduction to the Implementation of Automatic Differentiation
Yu-Hsueh Fang*, He-Zhe Lin*, Jie-Jyun Liu, and Chih-Jen Lin
Technical report, 2024
slides and code

A teaching-oriented walkthrough that implements a simple automatic differentiation system from scratch, bridging the gap between the math and a working implementation.

Places I've Been

The template is adopted from Jon Barron.