Zhishan Liu *** Taylor Ct
Mountain View, CA
Curriculum Vitae acn27j@r.postjobfree.com
Education
2010–2013 M.S. in Precision Instrument, Tsinghua University.
2005–2009 B.A. in Mechanical Engineering, Beijing Institute of Technology.
Courses
- Machine Learning, Introduction to Algorithms, Data Structure and Algorithm Analysis,
Probability Theory, Practical Common Lisp, Structure and Interpretation of Computer
Programs
- Familiar with common machine learning algorithms, such as Linear Regression, Bayesian
Classification, Gaussian Discriminant Analysis, K-means
- Familiar with common data structures and algorithms, such as Binary Search Tree, Graph,
Hash, Heap, Sorting, Divide and Conquer, Dynamic Programming
Experience
2013–2014 Meituan.com, Data Engineer, Machine Learning and Data Mining Team.
(Meituan.com is China’s version of Groupon.com)
Query Recommendation
- Designed and implemented all components for the recommendation service, mainly
including extracting featurs from log using ETL, training model, updating data to
redis, and writing service program, etc.
- Increased CTR(click through rate) by 20% through improving association rule algo-
rithm. The main idea was to treat association rule similarity as a feature, and then
use cosine similarity to find more similar queries.
- Increased CTR by 30% by using GBDT(Gradient Boosted Decision Trees) algorithm
to resort candidate queries.
- Skills: Python, ETL, Go, Redis, MySQL, Hive SQL, Thrift, Data Mining
Deal Recommendation
- Increased CTR by 15% through improving association rule algorithm to remove noise
in data, which could figure out the most probable deal branch based on users’ single-
branch-deal visiting history, and then remove multi-branch noise.
- Wrote a web application using tornado to show best cases and bad cases of deal
recommendation, which updated daily.
- Wrote different SQL to monitor performance of deal recommendation based on dif-
ferent standards, such as performance of cities, performance of classes, etc.
- Skills: Python, ETL, Redis, MySQL, Hive SQL, Recommendation System
Projects (https://github.com/liuzhishan/practice)
2014 Implemented PLSA(Probabilistic Latent Semantic Analysis) using Go
Given a set of articles, I used EM(expectation maximization) algorithm to compute model
parameters of PLSA, such as topic distribution and word distribution, and used goroutine
to compute parameters concurrently.
2014 Implemented decision tree using Go
Given a set of labeled data, I used mean square error as the regression criterion to build
decision tree. By presorting all samples according to one feature, the implementation
reached a O(Nf eatures Msamples (lgMsamples )2 ) time complexity.
2014 Solved 2-d k-nearest problem (Quora Programming Challenges) using kd-tree
The challenge asks programmers to find k-nearest points for every query, given N 2-d
points, k and Q query points(N and Q are in range [1, 10000]). I built kd-tree based on
given points and used priority queue to store nearest points, and then used the best-bin-
first method/recursive search method to look for the answer, which proved to be much
faster than the commonly used linear algorithm.
Skills
Domain Machine Learning and Data Mining
Languages Python, Go, C++, SQL, Common Lisp
Languages
English Be able to read English professional and technical books and take open courses without
difficulty.