Deep Learning Applications in Biological Pattern Recognition
Mentor 1
Dr. Peter Hinow
Location
Union Wisconsin Room
Start Date
5-4-2019 1:30 PM
End Date
5-4-2019 3:30 PM
Description
Interpreting experimental outputs takes up a large amount of lab time, and generally requires training in both the discipline as a whole as well as the experimental methods in particular. By automating parts of this interpreteation, work-hours can be better spent on other aspects of research, and results will be easier to verify by removing subjective judgements from the process. This study focuses on applying deep learning neural networks to the video output created by the water analysis experiments of Dr. Rudy Steiner. By using open source tools like Tensorflow and Python, various neural networks can be created and trained on existing data. Both fully connected and convolutional neural networks will be investigated, employing various network architectures in the search for the best solution. Efficiency, accuracy, and speed will be evaluated in the context of real time analysis on a small dedicated computer, using a raspberry pi. Preliminary programs are successfully classifying sample data, and it is expected that further refinement of models will yield faster and even more accurate results. If these techniques prove effective in analysis for Dr. Steiner’s work, it should be possible to generalise the technology to other types of water analysis that would benefit from reduced human interaction-- setting the stage for both better water safety testing as well as accelerated biological research.
Deep Learning Applications in Biological Pattern Recognition
Union Wisconsin Room
Interpreting experimental outputs takes up a large amount of lab time, and generally requires training in both the discipline as a whole as well as the experimental methods in particular. By automating parts of this interpreteation, work-hours can be better spent on other aspects of research, and results will be easier to verify by removing subjective judgements from the process. This study focuses on applying deep learning neural networks to the video output created by the water analysis experiments of Dr. Rudy Steiner. By using open source tools like Tensorflow and Python, various neural networks can be created and trained on existing data. Both fully connected and convolutional neural networks will be investigated, employing various network architectures in the search for the best solution. Efficiency, accuracy, and speed will be evaluated in the context of real time analysis on a small dedicated computer, using a raspberry pi. Preliminary programs are successfully classifying sample data, and it is expected that further refinement of models will yield faster and even more accurate results. If these techniques prove effective in analysis for Dr. Steiner’s work, it should be possible to generalise the technology to other types of water analysis that would benefit from reduced human interaction-- setting the stage for both better water safety testing as well as accelerated biological research.
