Please note that this course description is preliminary. The final course description will be released in June 2019.
- Bachelor's degree in Astronomy and/or Physics
- Fluency in Python and Linuxs
- Experience with processing big data sets
During the course students will be introduced to the field of Deep Learning and its possible applications to astronomy. Next, working in small groups, students will apply Deep Learning to some (open) problems.
- Basics of machine learning: classification, regression, clustering, overfitting, regularization
- Multi-layer Perceptron and Backpropagation, Stochastic Gradient Descent, Dropout
- Convolutional Networks for image classification
- Recurrent Networks for modelling sequential data
- Autoencoders for dimensionality reduction and extracting features from data
- TensorFlow, Keras and GPU-computing
- Learning basics of deep learning and its possible applications in astronomy
- Developing practical skills for designing and training deep networks
- Demonstration of newly acquired skills by solving some astronomy related problems
In this course, students will be trained in the following behaviour-oriented skills:
- Problem solving (recognizing and analyzing problems, solution-oriented thinking)
- Analytical skills (analytical thinking, abstraction)
- Creativity (resourcefulness, lateral thinking)
- Collaboration (extreme programming, joined research)
Mode of instruction
- Practical classes
- student's presentation
- Project (70%)
- Presentations (30%)
Lecture notes, additional readings and assignments will be provided through Blackboard.
For Blackboard access, you need an ULCN account. More information:
- Yann LeCun, Yoshua Bengio, and Geoffrey Hinton, Deep Learning, Nature 2015.
- Deep Learning, Yoshua Bengio, Ian Goodfellow, Aaron Courville, MIT Press (2017); available from: http://www.deeplearningbook.org/
- Keras documentation
- Selected papers from:
- Papers Reading Roadmap
- Papers Explained