Mohammad Ali Zamani

Mohammad Ali Zamani

Machine Learning Applied Scientist

HITeC

Biography

I am a Machine Learning Applied Scientist at Hamburg Informatics Technology Center (HITeC) and a Research Associate at University of Hamburg My research interests include Deep Reinforcement Learning, Computer Vision, Cognitive Robotics and Natural Language Processing.

I am passionate to find scalable solutions based on deep learning for industrial problems. My research focus has been on deep reinforcement learning including explainability and its applications in robotics, dialogue management systems, and planning. I am also interested in AutoML as a path to reach a scalable artificial intelligence solution.

Interests

  • Deep Reinforcement Learning
  • Computer Vision
  • Cognitive Robotics
  • Natural Language Processing

Education

  • Ph.D. Fellow in Computer Science

    University of Hamburg, Germany

  • M.Sc. in Computer Science, 2015

    Ozyegin University, Istanbul, Turkey

  • B.Sc. in Electrical Engineering (Major of Control Engineering), 2009

    University of Tehran, Iran

Experience

 
 
 
 
 

Machine Learning Applied Scientist

Hamburg Informatics Technology Center (HITeC)

July 2019 – Present Hamburg, Germany
  • Making workshops and tutorials about deep learning for internal and external audience.
  • Providing deep learning solution for our partner companies.
  • Developing deep learning (computer vision) methods to estimate device measurements from images.
  • Developing CNN models to detect defects on products from images.
 
 
 
 
 

PhD internship

IPA Fraunhofer

October 2018 – November 2018 Stuttgart, Germany
Developed a simulation compatible with deep reinforcement learning (DRL) for the rob@work robot in the Gazebo simulator.
 
 
 
 
 

Research Associate

University of Hamburg

May 2016 – April 2019 Hamburg, Germany
  • Developing state representation for DRL in action planning.
  • Developed DRL for visuomotor and dynamics in robots.
  • Developed low-latency continuous emotion recognition using DRL.
  • Improved robustness of emotion recognition on robots by incorporating spoken language and data augmentation.
  • Worked on personalized dialogue systems with dynamic preference using deep reinforcement learning.
 
 
 
 
 

Research Assistant

Ozyegin University

September 2012 – September 2015 Istanbul, Turkey
Introduced a human-robot skill synthesis framework based on simultaneous adaptation of the human and the robot skills. The human demonstrator learns to control the robot in real-time, at the same time, the robot learns from the human-guided control creating a non-trivial coupled dynamical system.
 
 
 
 
 

Research Assistant

University of Tehran

July 2009 – August 2012 Tehran, Iran

In cooperation with the Greater Tehran Electrical Distribution Company:

  • Developed expert system to adjust the level of automation using machine learning algorithms (especially, neural network).
  • Proposed decision-making approach for IT infrastructure selection in the smart grid.
  • Implemented a fault diagnosis in the smart grid using neural networks.

Recent Publications

Quickly discover relevant content by filtering publications.

Mixed-Reality Deep Reinforcement Learning for a Reach-to-grasp Task

Deep Reinforcement Learning (DRL) has become successful across various robotic applications. However, DRL methods are not …

Deep reinforcement learning using compositional representations for performing instructions

Spoken language is one of the most efficient ways to instruct robots about performing domestic tasks. However, the state of the …

Incorporating End-to-End Speech Recognition Models for Sentiment Analysis

Previous work on emotion recognition demonstrated a synergistic effect of combining several modalities such as auditory, visual, and …

Neural End-to-End Learning of Reach for Grasp Ability with a 6-DoF Robot Arm

We present a neural end-to-end learning approach for a reach-for-grasp task on an industrial UR5 arm. Our approach combines the …

On the Robustness of Speech Emotion Recognition for Human-Robot Interaction with Deep Neural Networks

Speech emotion recognition (SER) is an important aspect of effective human-robot collaboration and received a lot of attention from the …

Language-modulated Actions using Deep Reinforcement Learning for Safer Human-Robot Interaction

Spoken language can be an efficient and intuitive way to warn robots about threats. Guidance and warnings from a human can be used to …

Learning Spatial Representation for Safe Human-Robot Collaboration in Joint Manual Tasks

Programming robots for a safe interaction with humans is extremely complex especially in collaborative tasks. One reason is the …

Projects

SECURE Project

Safety Enables Cooperation in Uncertain Robotic Environments

Robot that performs language instructions

Presented at European Researchers’ Night in the Parlamentarium, Brussels, Belgium.

CONVERGENT Project

Human Learning for Robot Skill GenerationT

Teaching and Advising

Advising

  • Master Thesis: Adapting to User Context in a Reinforcement Learning-based Dialogue System by Mehran Sheikholeslami
  • Master Thesis: Reinforcement Learning for Incremental Dialogue Management by Thi Linh Chi
  • Master Thesis: Knowledge Extraction from API Reference Documentation Using Deep Learning by Alireza Mollaalizadehbahnemiri
  • Seminar Project: Deep Reinforcement Learning for Playing Games
  • Seminar Project: Playing Text-based games with Deep Reinforcement Learning
  • Seminar Project: Helping a Deep Reinforcement Learning Agent with Natural Language Instructions to Play a Video Game

TAship

  • Introduction to Machine Learning & Artificial Neural Networks, Ozyegin University, Spring 2013, Spring 2014, and Spring 2015.
  • Introduction to Programming (Matlab), Fall 2014
  • Discrete Mathematics, Ozyegin University, Fall 2014
  • Computer Programming (Java),Ozyegin University ,Spring 2014
  • Data Acquisition and Digital Control, Ozyegin University, Fall 2013,
  • Introduction to Robot Programming, Ozyegin University, Spring 2013
  • Physics Lab, Ozyegin University, Fall 2012, Fall 2014 Spring 2015
  • Electrical Machine laboratory, University of Tehran, Spring 2009-10
  • Programming for Mathematics and Statistics, Spring 2009
  • Electrical Machine I, University of Tehran, Fall 2009

Trainings and Summer Schools

  • Gaussian Process Summer Schools: Gaussian Process, Bayesian Optimization, Kernel Design, Deep GP, Sheffield, UK Sep. 2019

  • ACAI Summer School on Reinforcement Learning: Temporal Abstraction, Deep RL, Multi-Agent, Off Policy RLs, Nieuwpoort, Belgium Oct. 2017

  • Winter School on Humanoid Robot Programming: ROS, YARP, Kinematics, Dynamics, Vision, Santa Margherita Ligure, Italy Feb. 2017

  • Robot Workshop: ROS, LBR iiwa KUKA and Care-O-Bot programming, Stuttgart, Germany Dec. 2016

  • KUKA robot Training: KUKA Simulation Software, Istanbul, Turkey Jan. 2014

Awards

  • Gaussian Process Summer Schools Travel grant - MCAA EU Commission Sep. 2019
  • Marie SkIodowska-Curie Fellowship - SECURE Project 2016-2019
  • ICRA 2018 PhD forum travel grant award May 2018
  • Accepted project to “Science is Wonder-ful! - European Researchers’ Night” event Sep. 2018

Recent Talks and Workshops

Workshop on Deep Neural Learning and Bayesian Optimization of Hyperparameters

The (online) workshop is organized by Hamburger Informatik Technologie-Center e.V. (HITeC) and Artificial Intelligence Center Hamburg e.V. (ARIC) on 9th, 16th, and 23rd April 2020. The focus of the workshop was on optimizing hyperparameters of deep learning models with Bayesian optimization.

Workshop on Deep Learning with Pytorch and AutoML

My second internal workshop at HITeC was about deep learning with Pytorch and AutoML. I introduced the basics of deep learning and how to implement a classification problem in Pytorch. Then, I introduced one of the latest AutoML approach called BOHB for hyperparameter optimization.

Workshop on Gaussian Process and Bayesian Optimization

I had an internal workshop at HITeC on Introduction to Gaussian Process and Bayesian Optimization. The content of the workshop was: Basics from Statistics Gaussian Process Bayesian Optimization

2 poster presentations at IROS 2018

I jointly presented our paper On the Robustness of Speech Emotion Recognition for Human-Robot Interaction with Deep Neural Networks during the interactive section. Also, on the last day of the conference, I attended the Machine Learning in Robot Motion Planning workshop to present our work on Neural End-to-End Learning of Reach for Grasp Ability with a 6-DoF Robot Arm.

Presenting my research progress

We organized a PhD students conference focused safety and interaction quality. I also had a chance to present my progress in this conference.

Contact