A Course on "Theoretical aspects of machine learning" (SS 21)
Current Issues
- For any queries write an email to panda@na.uni-tuebingen.de.
- The first lecture was held on Tuesday (20th April 2021) via zoom.
- The interested students can register in the URM for the course or they can email to me.
Content of the Course
- Machine learning admits a very broad set of practical applications such as: Text or document classification, Natural language processing, Speech processing applications, Computer vision applications and fraud detection, etc. Most prediction problems found in practice can be cast as learning problems and the practical application area of machine learning keeps expanding. The techniques discussed in this course can be used to derive solutions for all of these problems.
- This course provides fundamental modern topics in machine learning while providing the theoretical basis, mathematical approach and conceptual tools needed for the discussion and justification of algorithms. We covered the following topics in our course:
- Learning theory; handwritten lecture notes
- The Probably Approximately Correct (PAC) learning framework; lecture notes
- Hoeffding inequality and Generalization bound;
- Growth function; lecture notes
- Vapnik-Chervonenkis-dimension (VC-dimension); lecture notes
- Complexities of the model;
- Bias-Variance analysis; lecture notes
- Overfitting and Underfitting;
- Regularization techniques; lecture notes
- Validation; lecture notes
- Stochastic gradient descent and Back propagation;
- Kernel methods;
Prerequisites
Basic knowledge in linear algebra, analysis and probability theory is needed. Some knowledge in elementary Hilbert space theory is also required (for Kernel methods).
Dates of Lectures
- The lectures were held every Tuesday (starting from 20th April 2021) from 4-6pm and conducted via Zoom.
References
- Foundations of Machine Learning, by M. Mohri, A. Rostamizadeh, A. Talwalkar, MIT Press, (2012)
- Understanding Machine Learning: From Theory to Algorithms, by S.S.-Shwartz, S.B.-David, Cambridge University Press, (2014)
- Statistical Learning Theory, by V.N. Vapnik, John Wiley & Sons, (1998)
Course Co-ordinator: Dr. Akash Ashirbad Panda