Mojtaba Taherkhani

Ph.D. Candidate in Smart computing

About Me

My M.Sc. thesis entitled: ’ Designing Large-Scale Recurrent Neural Networks with Incremental Learning’ broadened my knowledge in structure learning ,especially in dynamic environment. In my thesis, I worked on indirect coding to make large scale neural networks. My contributions include developing an evolutionary encoding to learn structure of NN, proposing a novel optimization with Stability-based Adaptive Inertia (SAIW) in order to learn the weights of neural network in dynamic environment, besides, I did research on how our model can learn the structure of helicopter controller during hovering maneuver in the simulation.

My proactive nature and desire to continue my research career made me work on different industrial domains such as financial projects and automotive projects which led me to deeply research on Deep Neural Networks (LSTM, GRU, GAN, CNNs, Deep Belief Network, Autoencoder). During my work, I have been developing some of these models with TensorFlow in PyCharm. In addition, these kinds of projects provided me an opportunity to develop some apps as backend developer by implementing distributed search engine (by Elastic), NoSQL in-memory data base (by Redis), Message-broker (by RabbitMQ), Business Intelligence panel (by Kibana) and data streamer (by Lightstreamer).

In my recent research, I am developing a deep reinforcement learning (RL) agent to model expert behaviors. I believe that I will be able to leverage the knowledge of this agent for scenario analysis challenges in digital twin applications.


Research Interests

Deep Neural Networks, Reinforcement Learning, Multiagent Learning, Graph neural networks,Proobabilistic Graphical Models, Behavioural Finance

Research Experience

Deep Learning Laboratory,Senior Research Scientist

Algorithmic Trading Department,Chief Data Scientist

CNN Group at Vision Laboratory,Machine Learning Researcher

Computational intelligence & Vision Laboratory,Research Assistant

Awards & Honors

Contact Information