Applied AI for Non-Programmers

11B1055

Bachelor

English

5.0

irregular

This module is offered as a block course within the International Summer University (ISU).

1 semester

As three week block seminar, this course will provide an overview of Artificial Intelligence (AI)-based on artificial neural networks (ANNs), i.e., feedforward neural networks, convolutional neural networks (CNNs), long short-term memory (LSTM) networks, and generative adversarial networks (GANs). In addition to theoretical concepts, the course covers practical aspects, including dataset design, evaluation metrics, programming languages, and common ANN frameworks. The students will perform a small ANN case study and present their results at the end of the course.

• main concepts of artificial intelligence (AI) focusing on artificial neural network (ANN)

• ANN zoo – discover the different neural network architectures

• ANN pipeline from data sets to applied ANNs via training of the ANNs

• ANN for classification tasks, time-series prediction, and continuous learning

• dataset creation

• pros and cons of AI and ANN

• evaluations figures of AI

• implementation of ANN with frameworks

The total workload for the module is 150 hours (see also "ECTS credit points and grading").

• Project Report, written

• Field work / Experimental work

The oral presentation of the project is done on-side during the block seminar. Approx. 3 weeks after the block seminar, the written seminar paper needs to be hand in digitally.

Project report (written): 20 pages on a completed practical project

experimental work: 4 task sheets, one each on feedforward neural networks, convolutional neural networks (CNNs), long short-term memory (LSTM) networks, and generative adversarial networks (GANs)

This course does not require any prior knowledge. It is mainly designed for undergraduate students who are majoring in non-informatic fields. Thus, this course is only open to non-informatic students.

Students who attend the course can know and understand different AI methods. With a focus on artificial neural networks, they can implement these methods with existing frameworks. Further, students should be able to:

• summarize the main concepts of artificial intelligence (AI) focusing on artificial neural network (ANN)
• describe the implementation pipeline from data sets to applied ANNs via training of the ANNs
• implement ANNs for classification tasks, time-series prediction, and continuous learning
• reate datasets (from real data) fitting the needs of AI
• know the challenges and limitations of ANNs
• understand the potential effects of AI on everyday life
• understand evaluations figures of AI
• distinguish whether or not the use of AI might outperform classical methods
• grasp issues based on unbalanced data set design, overfitting, underfitting as well as overgeneralization

Existing knowledge in the areas of designing algorithms and programming skills will be increased.

Students can (re) implement AI based on artificial neural networks in Python by frameworks such as TensorFlow, Pandas, and Numpy. They can also design and train superficial/shallow neural networks on existing data sets as a starting point for their own data sets and Tasks.

Students are capable of applying AI-based techniques and can estimate the training cost on existing datasets. They are also capable of selecting or combining suitable AI methods to prototype a specific task.

Students can (re) implement AI based on published code of artificial neural networks in Python by frameworks such as TensorFlow, Pandas, and Numpy. 

Students are capable of estimating the complexity and the applicability of AI methods and are capable of discussing these methods with peers. Further, they can give a summary of AI techniques to people from outside the field.

Students are capable of estimating the complexity and the applicability of AI methods and are capable of discussing these methods with peers.

• Schöning, Julius

• Schöning, Julius
• Tapken, Heiko
• Stiene, Stefan