Deep Learning for Neuroscience


Deep learning can achieve humanlike accuracy at tasks such as naming objects in a scene or recognizing optimal paths in an environment.  Sometimes it can even exceed human performance, recognizing non-obvious patterns in image or signal data.

In this new neuroscience seminar, we’ll illustrate the fundamentals of deep learning in MATLAB.  Using an age-labeled BIDS dataset from the OpenNeuro repository, we’ll train a deep network to accurately classify the age range of normalized human MRI brain images, not obviously discernable by human inspection. 


Along the way, participants will learn many aspects of the deep learning workflow:

  • Load and manage large sets of images
  • Import pre-trained models such as ResNet
  • Set up transfer learning via network modification
  • Get to network training quickly with apps for preprocessing and augmenting training image data
  • Configure network training parameters
  • Validation of convergence during deep model training
  • Interoperability with open source deep learning frameworks (i.e., TensorFlow-Keras, Caffe, PyTorch, etc.,) using ONNX
  • Accelerate algorithms on NVIDIA® GPUs, cloud, and datacenter resources without specialized programming.

Registration closed