Hire Sacha Dumay
Probability & Statistics Expert

Summary

After a double master degree from France and UK, my focus turned from robotics to data science. I have been working for the last 3 years for startup small and big and even one of my own as a data scientist. My expertise is in time series analysis, prediction and optimization. I love to work with data. That is why I also dived into backend development such as AWS, MongoDb and linux.

Subject Matter Expertise

• Probability & Statistics
• Forecasting & Prediction
• Machine Learning

Services

• Data Science

Work Experience

• Founder & Data Scientist

  NSJ Tech

  March 2017 - August 2018

• Data scientist

  Yello Mobile -Coocha

  June 2016 - February 2017

Education

• Master of Engineering

  ESTIA

  September 2014 - July 2016

• Master of Science

  University of Salford, Manchester

  September 2015 - June 2016

Publications

WORKING_PAPER
• {"ProfilePublicationId":43443,"Type":"WORKING_PAPER","Citation":"Abstract-This article presentsa new algorithm called GPMORL, “Generalised Probabilistic Multi Objective Reinforcement Learning”applied in target search.Probabilisticmethod allows the program to take in count uncertainty with stochastic values which is more than useful. Todiscretize more complex problem, scalarisation isused to combine several objective which droveto perform better.Main reinforcement learning such as Q-Learning and Dynamic Programming are explicitly applied to our application which is rescue and search with a drone.Finally, we demonstrate that the GPMORL can combine both assets of the different previous algorithms. Doing so, stability and convergence speed are improved significantly. Index Terms –Reinforcement Learning, SARSA, Prediction, Control, Q-learning,Gradient Descent, Temporal Difference error, Deterministic, Stochastic, Scalarisation, Multi-Objective, Dynamic Programming, Markov Decision Process \n\nGeneralised Probabilistic Multi Objective Reinforcement Learning | Request PDF. Available from: https://www.researchgate.net/publication/309314207_Generalised_Probabilistic_Multi_Objective_Reinforcement_Learning [accessed Sep 29 2018].","JournalTitle":"University of Salford, Manchester","Title":"Generalised Probabilistic Multi Objective Reinforcement Learning","CitationType":"FORMATTED_IEEE","DOI":"","Date":"2016-00-00","Url":"https://www.researchgate.net/publication/309314207_Generalised_Probabilistic_Multi_Objective_Reinforcement_Learning","PublicationContributors":null,"ContributorNames":"","TotalNumberOfPublications":0}
• {"ProfilePublicationId":43444,"Type":"WORKING_PAPER","Citation":"Abstract-People always care about their safety. Every government want to protect their citizens. But technology will partially resolve it. Indeed during the next decade self-driving cars will come to mass market. For that cars haveto followan itinerary while behavingsafely in any environment. Consequently, we investigate a mobile robotwhich could represent a self-driving car. This robot must be able to track someone while avoiding obstacles. It can also wandering around to potentially find someone to track. We have focused our work on a threshold logic unit to differentiate each behavior of our robot.In fact, we design this entire mechatronic system. We started to design the electronic design. Then we designed a three wheeled platform. Finally we implementedasimple artificial intelligenceand we tested our tracking by a T-Test. \n\nEmbedded Mobile Robot: TLU and Image recognition for safe tracking. | Request PDF. Available from: https://www.researchgate.net/publication/309314844_Embedded_Mobile_Robot_TLU_and_Image_recognition_for_safe_tracking [accessed Sep 29 2018].","JournalTitle":"","Title":"Embedded Mobile Robot: TLU and Image recognition for safe tracking.","CitationType":"FORMATTED_IEEE","DOI":"","Date":"0000-00-00","Url":"","PublicationContributors":null,"ContributorNames":"","TotalNumberOfPublications":0}