We spoke with Farzaneh Taslimi, Development Engineer and part of the Machine Learning Development team at MSC Software to discuss the popular topics of machine learning and artificial intelligence.
Can you tell me a bit about yourself?
I am Farzaneh Taslimi, an artificial intelligence developer and data analyst, with a great passion in program design and implementation. I received my PhD in computational soft matter physics from RWTH Aachen University, Germany. Prior to joining MSC Software, I was a postdoctoral scholar at the University of California, Irvine and a research fellow at Forschungszentrum Juelich, Germany.
What is your role at MSC?
Here at MSC Software, I am a member of the machine learning team. One of our projects is to develop simulation tools for training and validating autonomous vehicles, and to quantify the safety of the driving agents.
Why is machine learning important?
There is a quote which says, “Humans can typically create one or two good models a week; machine learning can create thousands of models a week”. Machine learning and data science are tremendously dynamic fields. Every day there are new subjects to learn. Because of the growing volume and varieties of available data, the demand for fast, powerful and cheaper computations are rising. This makes the field challenging and interesting at the same time.
Machine learning is a technique to train the computer to perform a task without being programed. Learning for a machine is somehow similar to human learning. It is actually improving the performance on a specific task by using statistical methods. We feed the machine with data on the task – training, and it learns by finding the trend of relation between features and target data.
What are some of the problems you are working on solving?
One of the projects we are dealing with is developing a simulator for training autonomous vehicles. Autonomous driving has attracted a lot of attention and interest in recent years and many car manufacturers have invested a lot of their resources to develop autonomous cars. Autonomous driving can reduce cost, crime, and increase mobility and user satisfaction.
An AV agent should have the ability to decide in an immediate event. Therefore, it needs to be trained with all possible scenarios that a vehicle can experience on real roads. Collecting real data to train the agent is too expensive and time consuming and it is almost impossible to generate all the possible scenarios. A simulator can help generate artificial data on a combination of possible scenarios to train the agent. We simulate the road environment by implementing VTD and generating different conditions on the road (number of cars, different maneuvers, weather conditions, etc.) by applying statistical tools.
One of the major concerns that play a dominant role in the development of autonomous vehicles is safety. With enormous improvement in autonomous vehicle performance, having a tool to calibrate safety and examine vehicles in different immediate scenarios is very important. Our mission is to raise the performance of autonomous vehicles to a higher level than a human driver by considering the errors in the sensors of the agents, different scenarios that an AI agent can experience, and the statistical tools.
Will AI and machine learning continue to affect the future of technology?
Absolutely! Increasing volume of the data and demand for predictive methods to analyze the data expands the machine learning application in different fields, from marketing purposes to designing medical devices. Applying AI and machine learning tools in technological applications can improve life quality, reduce production costs, and increase the speed of daily services.
What is your favorite part of what you do?
My favorite part is the day-to-day challenges we face and the research and creativity involved in solving them. It has me spending all day learning, testing and moving forward, expecting new challenges to come along the way.