Topic outline

  • Deep learning methods are based on a simple principle: extrapolate from previous examples. These methods fit non-linear functions, which are represented by artificial neural networks, to massive amounts of data. Such non-linear functions might then be used to detect the presence of a face mask (demo). Leveraging on increased availability of computation and data via the internet, deep learning methods achieve partially super-human performance within computer vision and natural language processing

    This course teaches you how to train deep neural networks on different kinds of data. The emphasis will be on image data as deep learning methods are most matured for image processing applications. We will not focus on details of certain calculations inside a deep neural network, but rather develop intuition and hands-on skills for applying deep learning methods. 

    The course material will be in the form of Python notebooks that contain code snippets along with their explanations. The notebook with accompanying material can be downloaded from https://github.com/alexjungaalto/DeepLearningPython.

    The course is heavily inspired by the excellent book "Deep Learning with Python" by F. Chollet, https://aalto.finna.fi/Record/alli.833878

    Workload of this course: 2 credits (approx. 60 hours of work); Grading is pass/fail.