Master's Thesis: Out-of-Distribution Detection Using - Bemidji, MN, United States - SilverLinx

SilverLinx
SilverLinx
Verified Company
Bemidji, MN, United States

3 weeks ago

James Miller

Posted by:

James Miller

beBee Recruiter
Freelance
Description

The Fraunhofer Society operates 76 institutes and research facilities in Germany and is the world's leading organization for applied research.

Around 30,000 employees work on an annual research volume of 2.9 billion euros.


Are you looking to gain experience in a modern and professional team and get an up-close look at how research and development in the field of human-machine interaction works? Then you've come to the right place At the Fraunhofer Institute for Optronics, System Technologies and Image Exploitation IOSB, we turn the latest scientific findings into technical innovations and want to contribute to shaping the future.


The distribution of training data for a neural network is crucial for the expected performance with a certain distribution of target input data.

If the data during the inference phase differs significantly from the training data, this can lead to unpredictable and potentially incorrect decisions by the network.

This is the case, for example, when the environment in which the network is used differs fundamentally from the environment from which the training data was generated.

Neural networks are often unable to recognize when their predictions may be incorrect due to unfavorable input data. Out-of-distribution detection determines whether the input data is within the range for which the network can provide reliable performance. It provides information on whether the output of the neural network is currently trustworthy or not. The aim of this work is to develop a module that evaluates to what extent the current input data of a neural network lies within the distribution of the training data
More jobs from SilverLinx
See all jobs of SilverLinx