Sami Khenisi

About Me

I am a Ph.D. student at the University of Louisville in Computer Science and Engineering, working at the Knowledge Discovery and Web Mining Lab and advised by Dr. Olfa Nasraoui.
My research revolves around Recommender Systems, specifically, I am trying to investigate problems such as Bias, Fairness, and Interpretability in state of the art Recommender Systems Practices

  • ResidenceUSA
  • AddressLouisville - KY

Research Interests

Machine Learning

As Machine Learning is taking over our daily activities leading to a large amount of Data being generated, I am interested in improving AI and ML algorithms. I am especially interested in Recommender Systems and its related topics

Bias, Fairness and Transparency in ML and AI

I am interested in investigating major flaws in ML algorithms related to Bias. For instance, I study feedback loop bias in Recommender Systems and their impact on both the user and the algorithm performance


Industry Experience


Research scientist

Research scientist within the VideoML org in Facebook

Working on advancing the state of Recommender systems for videos at Facebook

Summer 2021

Research Intern - Bloomberg AI Group

Project: Working with the AI Recommendations team on improving the Information Retrieval systems within Bloomberg. 

Summer 2016
Digisponsor - Paris

Data Scientist Intern

Project: Implementing an end-to-end tool to estimate the probability of success of crowdfunding projects using machine learning models.

• Designing a full data science pipeline for a crowdfunding french start up: Data collection, Data Analysis, Model Design, Model Evaluation and Launching into production.

Key words: Machine Learning, Data science, Data visualization, R, Python, Ruby, Microsoft Azure

Summer - 2015

Software Engineer Intern

Project: Develop a CRUD web application using PHP and SQL in order to manage a data center's equipment.

Key words: PHP, HTML, CSS, SQL

Research Experience

2017 - 2022
University of Louisville - Knowledge Discovery and Web Mining Lab

Graduate Research Assistant

Advisor: Dr. Olfa Nasraoui


Studying feedback loop bias in Recommender Systems
  • Theoretical modeling of the feedback loop in Recommender Systems
  • Studying bias in different training losses for common Recommender Systems Algorithms
  • Paper Accepted in Recsys 2020
Modeling and Counteracting Exposure Bias  in Matrix Factorization:
  • Inspecting the effect exposure bias on Matrix Factorization models.
  • Engineering new models for reducing exposure bias
  • Presented as Master Thesis at University of Lousiville
  • Received 1st place award at 2019 Speed Research Exposition Master Category
Designing a new explainable Active Learning Strategy for Recom-mender Systems
  • Explore explainability for Recommender systems model
  • Design a new Active Learning strategy for Matrix Factorization: ExAL Algorithm
  • Presented as Final Year Project In Tunisia Polytechnic School. Received the distinction Exceptional
  • Presented Poster in the Commonwealth Computational Summit - University of Kentucky


2019 - 2022
University of Louisville

Doctor of Philosophy - Ph.D, Computer Science

Advisor: Dr. Olfa Nasraoui

Awarded with the University Fellowhship

2018 - 2019
University of Studies

Master of Science - Computer Science

2014 - 2017
Ecole Polytechnique de Tunisie

Engineering Degree - Applied Mathematics

Graduation Project: New Explainable Active Learning Approach for Recommender Systems

Graduated with Distinction “Exceptional”

2012 - 2014
Institut Préparatoire aux études d'ingénieur de Tunis (IPEIT)

National engineering Entrance Exam Rank: 33/3000

Major: Mathematics and Physics

Top 1% in National competitive entrance exam to engineering schools: Admission at Tunisia Polytechnic School

Awards and Honors

  • University of Louisville Doctoral Fellowship 

    • University of Louisville – August 2019

  • 1st Place at 2019 Speed Research ExpositionMaster Category

    • University of Louisville – April 2019

  • Graduate Dean’s Citation

    • University of Louisville – April 2019

  • Highest cumulative scholastic standing in the departmental Program

    • University of Louisville – April 2019

  • Tunisian excellence scholarship

    • Tunisian Government – September 2014


Authors: Khalil Damak, Sami Khenissi, Olfa Nasraoui
Recent work in recommender systems has emphasized the importance of fairness, with a particular interest in bias and transparency, in addition to predictive accuracy. In this paper, we focus on the state of the art pairwise ranking model, Bayesian Personalized Ranking (BPR), which has previously been found to outperform pointwise models in predictive accuracy, while also being able to handle implicit feedback. Specifically, we address two limitations of BPR: (1) BPR is a black box model that does not explain its outputs, thus limiting the user’s trust in the recommendations, and the analyst’s ability to scrutinize a model’s outputs; and (2) BPR is vulnerable to exposure bias due to the data being Missing Not At Random (MNAR). This exposure bias usually translates into an unfairness against the least popular items because they risk being under-exposed by the recommender system. In this work, we first propose a novel explainable loss function and a corresponding Matrix Factorization-based model called Explainable Bayesian Personalized Ranking (EBPR) that generates recommendations along with item-based explanations. Then, we theoretically quantify additional exposure bias resulting from the explainability, and use it as a basis to propose an unbiased estimator for the ideal EBPR loss. The result is a ranking model that aptly captures both debiased and explainable user preferences. Finally, we perform an empirical study on three real-world datasets that demonstrate the advantages of our proposed models.

Authors: Sami Khenissi, Mariem Boujelbene, Olfa Nasraoui


The closed feedback loop in recommender systems is a common setting that can lead to different types of biases. Several studies have dealt with these biases by designing methods to mitigate their effect on the recommendations. However, most existing studies do not consider the iterative behavior of the system where the closed feedback loop plays a crucial role in incorporating different biases into several parts of the recommendation steps.
We present a theoretical framework to model the asymptotic evolution of the different components of a recommender system operating within a feedback loop setting, and derive theoretical bounds and convergence properties on quantifiable measures of the user discovery and blind spots. We also validate our theoretical findings empirically using a real-life dataset and empirically test the efficiency of a basic exploration strategy within our theoretical framework.

Authors: Wenlong Sun, Sami Khenissi, Olfa Nasraoui, Patrick Shafto

Abstract: Recommender Systems (RSs) are widely used to help online users discover products, books, news, music, movies, courses, restaurants, etc. Because a traditional recommendation strategy always shows the most relevant items (thus with highest predicted rating), traditional RS’s are expected to make popular items become even more popular and non-popular items become even less popular which in turn further divides the haves (popular) from the have-nots (unpopular). Therefore, a major problem with RSs is that they may introduce biases affecting the exposure of items, thus creating a popularity divide of items during the feedback loop that occurs with users, and this may lead the RS to make increasingly biased recommendations over time. In this paper, we view the RS environment as a chain of events that are the result of interactions between users and the RS. Based on that, we propose several debiasing algorithms during this chain of events, and evaluate how these algorithms impact the predictive behavior of the RS, as well as trends in the popularity distribution of items over time. We also propose a novel blind-spot-aware matrix factorization (MF) algorithm to debias the RS. Results show that propensity matrix factorization achieved a certain level of debiasing of the RS while active learning combined with the propensity MF achieved a higher debiasing effect on recommendations.






  • Modeling and Counteracting Exposure Bias In Recommender Systems, CSE Ph.D Seminar April 2019
  • Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System, Recsys October 2020
  • Understanding the Iterative Bias in Recommender Systems ML Tokyo October 2020

Volunteer Work

Summer 2018 and 2019
NSF Research Experience for Teachers in Big Data and Data Science at the University of Louisville

Graduate Student Mentor

As a part of the RET experience, I mentored a team of High School teachers through a 6-weeks program to help them with conducting research about Machine Learning. My responsibilities included providing python and programming courses, ML fundamentals, and guidance through the different project phases

Summer 2018:
- The team I mentored worked on explainability and recommender systems. They understood the basic models for RS and they implemented a collaborative filtering recommender systems and also provided explanation with the predictions

Summer 2019:
- I mentored two teams: The first team worked on explainable ML. They learned how to use explainability frameworks such as LIME with simple prediction tasks. The second team worked on Fairness in AI. They used the AI360 toolkit to investigate fairness for clustering algorithms on an income dataset.

Ecole Polytechnique de Tunisie

Founder of the Data Science Club at Ecole Polytechnique de Tunisie

Founder and president of the Data Science Club at Tunisia Polytechnic School
The club activities are:

  • Learning about Data Science pipeline through different lectures
  • Implement Machine Learning algorithms while understanding the theory behind
Ecole Polytechnique de Tunisie

Member of Enactus Tunisia Polytechnic School

Vice-champion Tunisia 2015
Winner of Innovation Award


  • Assist the company Grace Light Tunisia to implement a new solar air heater sales channel ECO-TECH.
  • Design a website for the team
Official Website for Ecole Polytechnique de Tunisie

Project Manager

Team leader and developer of the official website of  Ecole Polytechnique de Tunisie.
I managed a team of 10+ students to create a new website for Ecole Polytechnique de Tunisie. 


Machine Learning

  • Pytorch
  • Tensorflow
  • Deep Learning
  • Pytorch-Lightning
  • Keras
  • scikit-learn


  • Python
  • C/C++
  • Matlab
  • SQL
  • Java
  • R
  • Matlab
  • Go
  • Git
  • Bash

Big Data

  • AWS
  • Microsoft Azure
  • GCP
  • Spark/Pyspark
  • Slurm


  • Latex (Overleaf)
  • Markdown (Typora)
  • English
  • French
  • Arab



Predicting Estimated Repair Time for Power Outages for LG&E and KU Energy

Data Science, Machine Learning

Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System

Machine Learning, Recommender Systems, Research

Modeling and Counteracting Exposure Bias in Recommender Systems

Machine Learning, Recommender Systems, Research

An end-to-end intelligent assistance tool for crowdfunding projects

Data Science, Machine Learning, Software Engineering

Explainable Active Learning Strategy For Recommender Systems

Machine Learning, Recommender Systems, Research

CRUD web application in order to manage a data center’s equipment.

Software Engineering

Predict credit card Frauds using Anomaly Detection techniques (Kaggle competition – Silver Medal)

Data Science, Machine Learning

Implementing an Online Recommender system based on MovieLens data and ExplainableMatrix Factorization algorithm

Data Science, Machine Learning, Recommender Systems, Software Engineering

Implementing MapReduce for a word counting algorithm distributed across multiple AWS instances

Distributed Systems, Software Engineering


University of Louisville Louisville, KY 40292, USA

How Can I Help You?