Skip to main content

Practical Lab WS 18/19 Practical Lab Numerical Simulation

Algorithms in Machine Learning and Their Application

Under direction of
Prof. Jochen Garcke and Prof. Michael Griebel
Assistants
Dr. Bastian Bohn and Jannik Schürg
Date
Wednesdays, 14:15 to 16:00 in room 2.035 (Endenicher Allee 19b)
Tutorial
Wednesdays, 16:15 to 18:00 in room 2.038 (Endenicher Allee 19b)
Contact
Please use ed tod nnob-inu tod sni ta ballma tod b@foo tod de to contact us.

Material

To read the dataset for the river levels task, you may also use the Pandas DataFrame and read it with

import pandas as pd
df = pd.read_pickle("riverlevels.pandas.pickle")

Submissions to the exercise sheets

Each working group should send their submissions to the exercise sheets to ed tod nnob-inu tod sni ta ballma tod b@foo tod de. The solutions to the first sheet will not be discussed in detail with each group separately, but if you have certain questions or would like to check if you did things the right way, you can come to the tutorial dates and discuss your solutions with the tutor. The solutions to all other worksheets will be disussed with each group separately. Appointments will be made on short notice.

Content

In this practical lab, we teach the basic mathematical and technical tools needed to understand a range of basic data mining and machine learning methods. A strong emphasis is put on algorithms and efficient implementation.

Roughly every two weeks a new practice sheet is given to the participants. The tasks will be worked on in small groups. Depending on the technical proficiency, the time needed will be about 6 hours a week.

Background

Nowadays, data mining and machine learning algorithms are the backbone of decision making processes in all major enterprises. Their applicability seems almost endless and ranges from selective advertising over prototype design to autonomous production chains. Due to the availability of very large datasets (“Big Data”) it has become crucial to understand the mechanics of the different types of learning methods and to be able to develop and implement efficient algorithms to meet the requirements of the task at hand.

Requirements

Basic experience in Python is a necessary requirement. Further, the Python packages Numpy and Matplotlib will be used. The corresponding websites provide introductions which are sufficient for our purposes. All programming tasks are done using Jupyter notebooks. For the case of having no experience in the mentioned tools we recommened to spend a little time familiarizing yourself with these before the course starts.