Our workshop is titled “Introduction to Deep Learning and Creating Neural Networks (CNN and ANN) in Python and R.” The anticipated participants are those interested in deep learning and applying neural networks to biological datasets including medical imaging.
Recently, advanced computational methods and genomic deep learning techniques, such as neural networks, have become a powerful tool in bioinformatics. They enable computational biologists to work with large and complex medical datasets (DNA/RNA/protein sequences, medical imaging, etc.) that can be challenging to analyze with simpler machine learning models. It is with the addition of neural networks to our computational toolkits that we have been able to make predictions from big data (predicting protein secondary structure: disease diagnosis, subtype, and treatment etc.) and discover underlying biological principles not easily identified through using other algorithms. A team of researchers at the University of Minnesota recently developed an algorithm to evaluate chest X-rays to diagnose potential COVID-19 patients! (https://twin-cities.umn.edu/news-events/university-minnesotadevelops-ai-algorithm-analyze-chest-x-rays-covid-19).
The learning outcomes in this workshop include an introduction to deep learning and the various types of deep learning networks (ANN, CNN), how neural networks work, what they can be used for, as well as how to build a basic CNN and ANN in Python and R to assess biological data. A neural network is a machine learning framework that attempts to mimic the learning pattern of natural biological neural networks aka the human brain; therefore, it is designed to enable machines to solve problems humans can solve. Similar to how a biological neural network consists of interconnected neurons with axons that receive inputs and dendrites that produce outputs, an neural network consists of units called perceptrons or neurons that also receive inputs and produce outputs through nodes (there are potentially hidden nodes in between). Neural networks can be used with regression and classification problems, pattern recognition and detection. All demos and exercises will use a diverse set of biomedical data in python (TensorFlow API) and R.