Hands-on Introduction to Deep Learning

OHBM 2020 educational course

Deep Learning for Human Brain Mapping

Ariel Rokem, University of Washington

Follow along at: http://arokem.github.io/conv-nets-slides/


Machine learning

Learning from data

Features of the data are extracted/engineered

Training data is used to infer model parameters

Test data is used to evaluate accuracy

Artificial neural networks

A family of machine learning algorithms

Biologically inspired

Minsky and Papert (1969)

A cascade of linear/non-linear operations

Learn by back-propagating the errors

Learn by back-propagating the errors

For the top layer:

How would the error change if we changed the activity in each unit just a little?

For layer L-1

How much would change in each weight affect the activity in layer L?

Depends on:

  • How active is the node feeding into the weight?
  • How would a change in the weight affect the output of the node it feeds into?
  • How is the activity of this node affecting the error?
  • Stochastic Gradient Descent

    Choose a random batch from your training data

    Calculate the errors from the top of the network back

    Adjust the weights by a small amount

    Repeat until convergence

    Convolutional networks

    Inspired by the visual system

    Capitalize on spatial correlations in images

    MRI Quality control with DL

    Anisha Keshavan
    (UW eScience → Child Mind Institute → Octave Biosciences)

    Jason Yeatman
    (UW ILABS → Stanford)

    Quality of MRI images is a major bottleneck

    Let's get started!


    Contact information