Xiao (Cosmo) Zhang
I am an applied scientist at Amazon Web Service.


I obtained my Ph.D. degree in Computer Science from Purdue University in 2020, where I worked with Dan Goldwasser in the Natural Language Processing (NLP) group.


During my Ph.D. studies, my research interests were connectionism inspired machine learning methods for natural language processing, more specifically, neural networks and probabilistic models. I was particularly interested in structural models that try to understand language as a structured cognitive representation, focusing on the syntactic level of languages.


Before I started my Ph.D. studies, I worked on Neural Networks (Adaptive Resonance Theory) and numerical optimization methods for Human-Computer Interaction (HCI), with Gregory Francis in Psychological Sciences at Purdue University, from where I obtained my master in psychology (Mathematical and Computational Cognitive Science concentrated). During that period, I received intensive quantitative and computational training, and also developped my interest in machine learning.


Cosmo Zhang
Contact
{{spams}}
Information
Resume
Github
LinkedIn

Academic CountDown
{{details.name}}
{{typ.name}}:

{{paper.source}} ({{paper.short}}), {{paper.year}}

{{data.people[ky].Name}},

pdf abstract supplementary code data slides website
Some notes written by me in my own study or doing TA work.
IRWLS algorithm for MLE in logistic regression.  [pdf]
Distance between two hyperplanes.  [pdf]
I contribute to my own Github frequently.
Industry:
{{pos.title}}

{{pos.activity}}

With  {{data.people[ky].Name}},
{{pos.when}}

Teaching:

TA for Data Mining, 2016, taught by Bruno Ribeiro.
Head TA for Statistical Machine Learning, 2015, taught by Dan Goldwasser.
TA for Experimental Design and ANOVA, 2015, taught by Richard Schweickert.
TA for Statistical Inference, 2014, taught by Richard Schweickert.
Head TA for Introduction to the Analysis of Algorithms, 2013, taught by Mike Atallah.
TA for Introduction to Cognitive Psychology, 2011 and 2012, taught by Gregory Francis.


Purdue University
Ph.D. in Computer Science, 2020
Dissertation: Flexible Structured Prediction in Natural Language Processing with Partially Annotated Corpora
Advisor: Dan Goldwasser

Purdue University
Joint Master in Computer Science and Statistics, 2015
Advisors: Jennifer Neville, Dan Goldwasser

Purdue University
Master in Psychological Sciences (with thesis), 2014
Thesis: Optimization of Switch Virtual Keyboard by Using Computational Modelling
Advisor: Gregory Francis

Huazhong Normal University, P.R.China
Bachelor of Science in Psychology, 2011

Huazhong Agricultural University, P.R.China
Bachelor of Arts in Linguistics, 2011

In the graduate school, I have taken the following relevant courses (overall GPA 3.7):
Computational Methods in Optimization, Operating Systems, Machine Learning Theory, Statistical Machine Learning, Advanced Statistical Methods, Machine Learning Methods for Natural Language Processing, Database Systems, Bioinformatics Algorithms, Introduction to Mathematical Statistics, Numerical Linear Algebra, Introduction to Cognitive Neuroscience, Computational Statistics, Data Mining, Algorithm Design and Analysis, Neuromorphic Systems and Vision, Introduction to Probability, Bayesian Applied Decision Theory, Introduction to Simulation and Modeling of Computer Systems, Vector Calculus
Some useful links I would like to archive.
Deep Learning website

Andrew NG's lecture notes for Machine Learning (Stanford CS229)

Good reference website for Latex

A very decent reference website for probability and statistics
I like reading books during my spare time. I am also fond of hiking, travelling and photography. I served as the president of Purdue Computer Science Graduate Student Board from 2013 to 2014, having organized activities and connected graduate students to industrial companies as well. I am a hard-headed behaviorist. B. F. Skinner is my favourite psychologist. Though in psychology it is widely claimed that behaviorism has been substituted by cognitive psychology, I believe the essence of behaviorism was successfully inherited by machine learning.