06.01.2022
| Inside Data Science |
Christopher Metz über seine Forschung am DSC
DSC Experte Christopher Metz will mit seiner Forschung die Effizienz von neuronalen Netzen verbessern. Im Interview spricht er über maschinelles Lernen, die richtige Infrastruktur und das Problem der Datenverfügbarkeit.
What topics are you currently working on in your research?
I am interested in building data science and machine learning systems. One major challenges is to pick the right device for applying trained AI models. It becomes critical in the areas of edge computing and IoT (Internet of Things) were tight power and performance constrains exists.
Furthermore, I am also interested in the topic of High-Performance Computing (HPC) systems for training such algorithms. I always try to improve the infrastructure of the Data Science Center. However, in my PhD research I am focusing more in the power and performance estimation on edge and IoT devices than on the High-Performance Computing for training.
How important is data to your research?
In general data is very important for me because I am using most time machine learning methods. Unfortunately, there are no public research data I can use. Hence, I need to create my own training data. This is often a time intensive process. I am looking forward to the development of open data and FAIR research data management to solve this kind of issues.
What role does data science play in your research? Do you see yourself more as a user, a method developer, a basic researcher, or perhaps something completely different?
Most time I am using machine learning methods which are a subclass of data science techniques. So, I would say data science is important to my research. Moreover, I am working on hardware and IT architectures for data science/ML.
I would call myself more of a Data/System Engineer who builds infrastructures and supportive platforms for data scientists.
Which data science methods and technologies are in the focus of your research or could also become interesting in the future?
Currently, I am focusing on power and performance estimation for neural networks on different devices. This makes – obviously – neural networks important to me. Moreover, I am applying methods like regressions or clustering methods on my data. This makes algorithms like K-Means, K-Nearest Neighbors and many more also important.
What are your main challenges in dealing with data?
My main challenge is to get the data. Due to the fact, that there are no available public data I need to execute benchmarks on different devices and create my own data. This is time and cost intensive. We do not have all devices available and it is not manageable to buy a device only for this purpose.
Another challenge is to get the data into a good shape and find the important features for training. We can collect dozens of attributes during our benchmarks but do not know which have a high impact to our output.
And finally, what is your personal motivation for joining the Data Science Center?
I applied for the role at the Data Science Center because I saw the opportunity to build a whole new platform and infrastructure for researchers. In my former job, I have always told people we need a group of experts to handle infrastructure and services for data science/AI applications for other researchers who are not into computer science. Since this is one of the things we do at DSC, I am very happy to be a part of it for almost two years now!
Furthermore, the DSC offers the opportunity to work on my PhD thesis which was not that easily possible at my former job.
I am looking forward to work in interesting and highly innovative projects at the DSC with all the different researcher of the Uni Bremen.
You can learn more about Christopher’s activities and the DSC infrastructure in his talk
“Do you pick the right device for data science?“ in the Data Science Forum on 13.01.2022.
« zurück
06.01.2022 | Inside Data Science
Christopher Metz über seine Forschung am DSC
DSC Experte Christopher Metz will mit seiner Forschung die Effizienz von neuronalen Netzen verbessern. Im Interview spricht er über maschinelles Lernen, die richtige Infrastruktur und das Problem der Datenverfügbarkeit.
What topics are you currently working on in your research?
I am interested in building data science and machine learning systems. One major challenges is to pick the right device for applying trained AI models. It becomes critical in the areas of edge computing and IoT (Internet of Things) were tight power and performance constrains exists.
Furthermore, I am also interested in the topic of High-Performance Computing (HPC) systems for training such algorithms. I always try to improve the infrastructure of the Data Science Center. However, in my PhD research I am focusing more in the power and performance estimation on edge and IoT devices than on the High-Performance Computing for training.
How important is data to your research?
In general data is very important for me because I am using most time machine learning methods. Unfortunately, there are no public research data I can use. Hence, I need to create my own training data. This is often a time intensive process. I am looking forward to the development of open data and FAIR research data management to solve this kind of issues.
What role does data science play in your research? Do you see yourself more as a user, a method developer, a basic researcher, or perhaps something completely different?
Most time I am using machine learning methods which are a subclass of data science techniques. So, I would say data science is important to my research. Moreover, I am working on hardware and IT architectures for data science/ML.
I would call myself more of a Data/System Engineer who builds infrastructures and supportive platforms for data scientists.
Which data science methods and technologies are in the focus of your research or could also become interesting in the future?
Currently, I am focusing on power and performance estimation for neural networks on different devices. This makes – obviously – neural networks important to me. Moreover, I am applying methods like regressions or clustering methods on my data. This makes algorithms like K-Means, K-Nearest Neighbors and many more also important.
What are your main challenges in dealing with data?
My main challenge is to get the data. Due to the fact, that there are no available public data I need to execute benchmarks on different devices and create my own data. This is time and cost intensive. We do not have all devices available and it is not manageable to buy a device only for this purpose.
Another challenge is to get the data into a good shape and find the important features for training. We can collect dozens of attributes during our benchmarks but do not know which have a high impact to our output.
And finally, what is your personal motivation for joining the Data Science Center?
I applied for the role at the Data Science Center because I saw the opportunity to build a whole new platform and infrastructure for researchers. In my former job, I have always told people we need a group of experts to handle infrastructure and services for data science/AI applications for other researchers who are not into computer science. Since this is one of the things we do at DSC, I am very happy to be a part of it for almost two years now!
Furthermore, the DSC offers the opportunity to work on my PhD thesis which was not that easily possible at my former job.
I am looking forward to work in interesting and highly innovative projects at the DSC with all the different researcher of the Uni Bremen.
You can learn more about Christopher’s activities and the DSC infrastructure in his talk
“Do you pick the right device for data science?“ in the Data Science Forum on 13.01.2022.
Interview Partner:
Christopher Metz, M.Sc.
Wissenschaftlicher Mitarbeiter am DSC
FB 03 – Mathematik und Informatik
cmetz@uni-bremen.de
« zurück