Data Scientist Colin Flynn explains deep learning in about 90 seconds.
www.dewberry.com
Hi, I'm Colin Flynn and I do machine learning and deep learning research at Dewberry's geospatial and technology services group. Deep learning's a subfield of machine learning that tries to mimic how our brains work and how they learn. So, it's using artificial neural networks and layers of many, many individual neurons that each have their own adjustable weights and biases that allow for the models to actually learn through a process called backpropagation. As a branch of machine learning, deep learning requires less human input and actual time from the human to get going and to achieve the desired results. Deep learning models can take different types of input data such as imagery, audio, anything that can be turned into a number that the model can then read in and use as an input.
For example, Dewberry is using aerial imagery to support post-disaster damage assessments. So, a model trained on aerial imagery could input new images and make an assessment based on the actual pixels within the imagery. For this example, the prediction could be that the buildings are damaged major, damaged minor undamaged. But deep learning models can be trained to address a variety of different problems.