The »Self-Learning Systems« group at Fraunhofer IIS is part of the »Precise Positioning and Analytics« department in Nuremberg. Our team members have diverse academic backgrounds from computer science, engineering and physics.
One of our research goals is to provide robust and safe algorithmic solutions for control and decision-making tasks in complex dynamic environments. To this end, we employ a range of approaches from machine learning and deep learning in combination with reinforcement-learning techniques. This allows us to generate adaptive decision-making policies capable of processing high-dimensional state representations, such as camera images and other types of sensor readings.
In recent years, the use of deep neural networks has vastly increased. However, when used on mobile and embedded devices, the model complexity and the size of updates become important factors. Thus it is not surprising that the field of »Deep Compression« has emerged to algorithmically reduce the size of models of neural networks. This can be used to drastically reduce the size of software updates for, e.g., smart devices. Even more recently, algorithms for »Distributed Learning« were introduced. Both methods aim at reducing the bandwidth required for application of neural networks. As in many applications the systems themselves need to adapt, we want to discover the best strategy to optimize for both network bandwidth and overall system performance of a self-learning system.
Your tasks: You …
- will implement and compare deep compression and distributed learning algorithms
- will learn to work with deep compression frameworks (Nervna distiller, NEMO, PyTorch, …)
- will test adaptive methods on different use cases form industrial tools to self-localization and self-navigation
What we expect from you
- You are currently enrolled in a physics, mathematics or computer science program
- You have some machine learning background and have already worked with machine learning frameworks
- You are interested in deep compression and/or distributed learning
- You speak English and/or German fluently
What you can expect from us
- An open and cooperative working environment
- Collaboration in interesting and innovative projects
- Many opportunities to gain practical experience and attend seminars
- Flexibility concerning your working hour
Applications are possible in German and English. Please include a cover letter, your CV and your latest transcripts of records (as PDF) and quote ID number 52748-LV. Address your application to Nina Wörlein.
Please let us know how you learned about this job opportunity.
|Job Reference: 52748-LV||Closing Date:|