Profile image of ninoarsov
@ninoarsov
Member since January 28, 2013
0 Recommendations

ninoarsov

Online Offline
I am working as a machine learning researcher at the Macedonian Academy of Sciences and Arts. Previously, I was a visiting researcher at Stanford University where I worked with the SNAP group (part of the InfoLab). There, I worked on multiple machine learning and data science projects. I have over four years of research experience in machine learning. Extensive knowledge of Python (including various ML libraries), C/C++, Matlab, etc.
$10 USD/hr
2 reviews
1.5
  • 100%Jobs Completed
  • 100%On Budget
  • 100%On Time
  • N/ARepeat Hire Rate

Portfolio

Recent Reviews

  • image of Adelaide L. Matlab expert with lower rate $30.00 SGD

    “He is very helpful and very patient. When my codes did not worked out the first time, he would look into it and resolved it.”

  • image of Vamsi Manikanta S. technical writer ₹2500.00 INR

    “Highly Recommend.”

Experience

Machine Learning Researcher

Jan 2019

I am working as a machine learning researcher on various projects related to network science, mining of large graph data, and machine learning in general. My projects involve the stability of machine learning algorithms, machine learning applications in finance, and Bayesian analysis. All of the projects involve extensive data preprocessing and preparation as well as visualization. You can find my papers at [login to view URL] and [login to view URL]

Visiting Researcher

Sep 2016 - Dec 2016 (3 months)

Visiting researcher supervised by Prof. Jure Leskovec ([login to view URL]) who is heading the SNAP group in the InfoLab at Stanford University. I worked on various research projects in machine learning, including unsupervised node embedding of real-world massive networks, mining of massive gene expression datasets for cancer research, mining of employee databases from the Royal Bank of Canada, helping humans understand why and how machine learning models fail to generalize.

Volunteering Researcher

Jan 2015 - Apr 2017 (2 years)

Volunteered at the Laboratory for Complex Systems and Networks Research (now called Research Center for Information Technologies). All projects involved developing new machine learning algorithms.

Database Engineering Intern

Jul 2014 - Aug 2014 (1 month)

Worked on implementing of the TPC-C Benchmark in VoltDB (distributed database system) using Java and setup of a VoltDB distributed experimental environment in FCSE’s Nebula Cloud infrastructure.

Database Research Intern

Jul 2013 - Sep 2013 (2 months)

In collaboration with FCSE, Skopje – Implementation of a genetic algorithm in C++ for column-store database compression by run-length encoding using heuristic column-reordering approaches. Additionally, byte-aligned variable-length encoding was implemented to optimize compression ratios.

Volunteering Undergraduate Researcher

Jun 2013 - Jun 2015 (2 years)

Implemented a genetic algorithm and other metaheuristics to find the optimal order and sort direction of columns in a column-store table in order to maximize RLE compression rate in database systems. [login to view URL] Developed a new machine learning algorithm in which boosting ensembles collaborate among each other within a bagging ensemble in order to minimize the generalization error of this two-level complex ensemble. To read more: [login to view URL]

Tutoring students

Jan 2011 - Dec 2015 (4 years)

Programming-related courses, Calculus, Probability and Statistics, Discrete Mathematics.

Education

Bachelor of Science (GPA: 9.74/10.00)

2011 - 2015 (4 years)

Master of Science (Current GPA: 10.00/10.00) [expected June 2019]

2018 - 2019 (1 year)

Josip Broz Tito High School (Mathematics) GPA: 5.00/5.00

2007 - 2011 (4 years)

Qualifications

Best Student Award (2016)

Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University

Best Student Award (2015)

Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University

Best Student Award (2014)

Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University

Awarded for achieving a total GPA of over 9.50 out of 10.00.

'Best IT Students in the Republic of Macedonia' Scholarship (2014)

Government of the Republic of Macedonia

Awarded to top Computer Science students in the country.

'Best IT Students in the Republic of Macedonia' Scholarship (2013)

Government of the Republic of Macedonia

Awarded to top Computer Science students in the country.

'Best IT Students in the Republic of Macedonia' Scholarship (2012)

Government of the Republic of Macedonia

Awarded to top Computer Science students in the country.

Best Student Paper (2014)

Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University

First author of a paper chosen the best among 66 papers, part of the Research Methodologies in ICT course.

Publications

Generating highly accurate prediction hypotheses through collaborative ensemble learning

[login to view URL] Supplemenatry Information (PDF): [login to view URL] First author. In this paper, we propose a novel collaboration approach between Gentle Boost ensembles. We use Bagging to combine the boosting ensembles and collaboration is driven by stability theory. We provide extensive mathematical proofs to support our approach. On average, our methods were able to decrease the generalization error of the models by 40%.

Generalization-aware structured regression towards balancing bias and variance

[login to view URL] In this paper, we present a novel bias-variance balancing objective function is introduced in order to improve the generalization performance of ensembles for structured regression. Our method is called Generalization-Aware Collaborative Ensemble Regressor (GLACER). GLACER was ∼10-56% and ∼49-99% more accurate for the task of predicting housing prices and hospital readmissions, respectively.

A meta-heuristic approach for RLE compression in a column store table

[login to view URL] In this paper, we present a genetic algorithm for finding the optimal order of columns in a column-store table in order to maximize the compression rate of the Run-Length encoding (RLE) compression algorithm. We analyze other heuristic algorithms, such as simulated annealing, cuckoo search, particle swarm optimization, Tabu search, and the bat algorithm. Our algorithm is applicable to large column-oriented database systems.

Stacking and stability

Preprint available at: [login to view URL] In this paper, we analyze the stability of the Stacking algorithm. We algebraically prove that Stacking improves Bagging, and vice versa. The stability can be leveraged to derive and optimze upper bounds on the generalization error of any machine learning algorithm.

Stability of decision trees and logistic regression

Preprint available at: [login to view URL] In this paper, we derive the stability of Decision Trees and Logistic Regression for the first time. We prove that deeper decision trees are more stable. We also show that the stability of logistic regression is not controllable unless the algorithm is regularized beforehand. Our results can be used to improve these algorithms by tightening the stability-based upper bound on their generalization error.

Collaborative bagging of boosting ensembles

[login to view URL] This is a poster presented at the South-East European Forum on Data Science in Belgrade, Serbia, 2016. It shows how collaboration between boosting ensembles reduces the error rate.

Weighted Bagging Predictors

[login to view URL] This poster presents an approach to learn the weights of the members of a bagging ensemble. The goal is to further reduce the error rate of the ensemble.

Certifications

  • US English Level 1
    82%

Verifications

  • Facebook Connected
  • Preferred Freelancer
  • Payment Verified
  • Phone Verified
  • Identity Verified
  • Email Verified

My Top Skills

Browse Similar Freelancers