I'm Johann Schmidt
I'm a Scientist.
I'm a Researcher.
I'm a Developer.
Welcome
based in Magdeburg, Germany.
Get in contactAbout Me
Let me introduce myself
I'm Johann Schmidt, PhD Student in Deep Learning
I work in the AILab of the Otto-von-Guericke University in Magdeburg, Germany as a researcher, where my daily business are deep neural network. My research focuses on fortifying these networks against vulnerabilities caused by variances in the input.
Take, for instance, an image classifier. It will break if issued with a rotated replicate of the training data. I develop solutions using the immense power of group theory. Join me on my quest to make neural nets exponentially more powerful and stable.
- Name:Johann Schmidt
- Email:johann.schmidt@ovgu.de
- Age:28
- From:Magdeburg, Germany
10+
Years Coding Experience
20+
Students Supervised
100+
Sessions Moderated
8
Papers Published
Career
Resume
Education
2007 - 2014
A Levels
Privatgymnasium Tangermünde, Germany
During this shaping period, my fascination with electrical circuits, programming and math emerged, igniting a passion that continues to fuel my academic and professional pursuits.
2014 - 2018
Bachelor: Electrical Engineering
University of applied Science Magdeburg-Stendal, Germany
Although adept at designing electrical circuits, my true passion was unveiled in computer science - breathing life into these circuits.
Bachelor Thesis
Institut for Automation and Communication (ifak) e.V. Magdeburg, Germany
I developed a GUI for a clamp-on ultrasonic sensor system on an embedded controller in C, as well as analysed and evaluated the overall measurement system.
2018 - 2020
Master: Digital Engineering
Otto-von-Guericke University Magdeburg, Germany
I delved into advanced computer science courses, dedicating countless hours to self-study to bridge gaps in my knowledge. It was within the realm of deep learning that I found my niche—a domain where I finally felt at home.
Master Thesis
Fraunhofer Institute for Factory Operation and Automation IFF Magdeburg, Germany
I developed and integrated a vision-based human action recognition deep learning model for human-machine interaction in an industrial environment.
2020 - today
PhD in Deep Learning
Otto-von-Guericke University Magdeburg, Germany
I cultivated a robust expertise in deep learning, immersing myself in diverse research domains. I mentored students, refined scientific writing through conference publications, delivered presentations, and adeptly facilitated and guided small groups, cementing both theoretical and practical facets of the field.
PhD Thesis
AILab, Otto-von-Guericke University Magdeburg, Germany
I invent algorithms to efficiently search the Affine Group and canonicalize input signals of Deep Neural Networks boosting their stability when issued with out-of-distribution data.
Experience
2013 - 2016
Internships and Student Jobs
TCS AG Genthin, Bachmann GmbH Magdeburg, Enercon GmbH Magdeburg
I first came in touch with the development and assembling of electrical circuit boards. I analysed and evaluated of wind turbine data and helped with cabling of control cabinets of wind turbines.
2020 - 2022
SENECA
BMBF-funded Project
During this BMBF-funded project we supported a regional semi-conductor company (Tectron Worbis GmbH) by developing an AI to tackle an NP-hard scheduling problem.
2022 - 2022
AI-Engineering
BMBF-funded Project
We created a novel Bachelors study program teaching the theories of artificial engineering intertwined with practical applications in several engineering disciplines (hands-on and student-centred with OERs).
2022 - today
PASCAL
BMBF-funded Project
Currently, we are lifting traffic control to the 21st century by developing an AI-based traffic signal control system using real-time V2X data.
Papers
Publications
2021
Human Action Recognition as part of a Natural Machine Operation Framework
S. Bexten, J. Schmidt, C. Walter, and N. Elkmann
IEEE International Conference on Emerging Technologies and Factory Automation (ETFA)
2021
NeuroEvolution of augmenting topologies for solving a two-stage hybrid flow shop
S. Lang, T. Reggelin, J. Schmidt, M. Müller, and A. Nahhas
Expert Systems with Applications
2021
Approaching Scheduling Problems via a Deep Hybrid Greedy Model and Supervised Learning
J. Schmidt and S. Stober
IFAC Symposium on Information Control Problems in Manufacturing
2024
Reviving Simulated Annealing: Lifting its Degeneracies for Real-Time Job Scheduling
J. Schmidt, B. Köhler, H. Borstell
The Hawaii International Conference on System Sciences (HICSS)
2024
Pursuing the Perfect Projection: A Projection Pursuit Framework for Deep Learning
J. Perschewski, J. Schmidt, S. Stober
Conference on Artificial Intelligence and Statistics (AISTATS)
2024
Learning Continuous Rotation Canonicalization with Radial Beam Sampling
J. Schmidt and S. Stober
under review
2024
Tilt your Head: Activating the Hidden Spatial-Invariance of Classifiers
J. Schmidt and S. Stober
under review
Blog
Little Knowledge Chunks
It's a relationship between two shapes or spaces that preserves their essential structure. Imagine you have a rubber band and a circle. A homeomorphism would mean you can stretch, twist, or bend the rubber band to turn it into a circle without tearing it. They might look different, but they have the same fundamental properties.
In mathematics, particularly in algebra or structures like groups or graphs, an isomorphism is a way to show that two different things are fundamentally the same. For example, if you have two graphs that have the same connections between nodes, just labeled differently, they are isomorphic. It's like having two jigsaw puzzles with different pictures but the same shapes of pieces.
An automorphism is a specific type of isomorphism. It's like a transformation that something can undergo and still remain itself. In mathematical terms, it's an isomorphism from a mathematical object to itself. For instance, in a square, rotating it by 90 degrees is an automorphism because it's still the same square.
In the realm of calculus and differential geometry, a diffeomorphism is a special type of mapping between two differentiable manifolds. It's a transformation that's smooth and has a smooth inverse. Imagine crumpling a piece of paper and then unfolding it back perfectly; if this transformation is smooth and reversible, it's a diffeomorphism.
Contact
Get in Touch
Address
Otto-von-Guericke University
FIN, IKS, AILab
Universitätsplatz 2
39106 Magdeburg, Germany
+49 391 67-54907
johann.schmidt@ovgu.de