Machine Intelligence II (unsupervised methods)

SoSe22

• Details regarding the weekly format will be explained in the ISIS course
• The ISIS course will be opened shortly before the lecture period starts. Access to the ISIS course requires a tubIT account.
• The lecture and tutorial will be offered in-person (dt. in Präsenz).
• 1 lecture and 1 tutorial every week. No need to attend more than one tutorial during the same week.
• We don't take attendance.
• Access to recorded video material from earlier years will be made available in case you missed a lecture.
• The exam(s) will take place on campus (dt. Präsenzklausur). You will need to be in Berlin to take part in the exam.

General information

• The courses Machine Intelligence I and II can be heard independently.
• Machine Intelligence II is offered in the summer while Machine Intelligence I is offered in the winter.
  
• Information regarding the material, tutorials and the exam can be found on the ISIS page. See the link to our ISIS course on the panel to the right. ⟶
• The lecture and tutorials are held in English.
• The course is open to TU students as well as exchange students and visiting students. Non-TU students should apply for a Neben-/Gasthörschaft to gain access to the course material.
• You do not need a formal registration to attend the course. You can self-enroll to the ISIS course as soon as its available. The registration is only relevant for the exam in order to earn ECTS points.
• The exam registration procedure will be explained in the ISIS course. Prior registration/reservation is not possible and not necessary.

Topics covered

• Principal Component Analysis
• Hebbian learning
• Kernel PCA
• Independent Component Analysis
• Stochastic optimization
• K-means clustering
• Pairwise clustering
• Self-Organizing Maps
• Locally Linear Embedding
• Probability density estimation
• Mixture models & Expectation-Maximization algorithm
• Hidden Markov Models
• Estimation theory

Prerequisites

• Solid mathematical knowledge: analysis, linear algebra, probability calculus and statistics. We emphasize this requirement because the course deals with the theoretical aspects and mathematical formulations of the learning algorithms.
• Basic programming skills, preferably Python, R, Matlab, or Julia. The programming skills are relevant for solving the programming exercises.

Target Audience / Assessment and Grading

For further information please consult Klaus Obermayer (lecturer, Modulverantworthlicher/in-charge) or Youssef Kashef (assistant).