TU Delft
Year
NEDERLANDSENGLISH
Organization
Education Type
Education print this page print this page     
2017/2018 Electrical Engineering, Mathematics and Computer Science Master Computer Science
CS4065
Multimedia Search and Recommendation
ECTS: 5
Responsible Instructor
Name E-mail
Prof.dr. A. Hanjalic    A.Hanjalic@tudelft.nl
C.C.S. Liem    C.C.S.Liem@tudelft.nl
Contact Hours / Week x/x/x/x
0/0/0/4+lab
Education Period
4
Start Education
4
Exam Period
4
5
Course Language
English
Course Contents
Nowadays, a huge amount of multimedia data is available online. While this has the potential to serve a multitude of use cases, the sheer amount and diversity of available multimedia data and consumer information needs require the development of sophisticated access mechanisms. Furthermore, the term "multimedia" implies that user queries and data to be handled are rich and multimodal (combining text, image, video, audio, etc).

In this course, methods, algorithms and best practices are discussed which deploy this richness of information to maximize the effectiveness, efficiency and intuitiveness of multimedia search and recommendation. Furthermore, implications of the fact that the data is consumed in networked communities of human users are treated.

After three weeks of core topics, the course offers two specialization tracks:
• MMSR Analytics, focusing on data analytics aspects for multimedia search and recommendation with special focus on emerging topics.
• MMSR Systems, focusing on system and implementation aspects for multimedia search and recommendation with special focus on handling real-world multimedia data.
Study Goals
Students will be able to
• explain the concept of “multimedia”;
• explain the principles underlying basic multimedia search engines;
• explain the functioning of basic multimedia recommender systems;
• describe and implement common representations of multimedia content;
• describe and implement common ranking mechanisms for multimedia search;
• describe and implement common recommender system techniques;
• describe and implement common social media analytics techniques for multimedia search and recommendation;
• interpret current academic literature in the field of multimedia search and recommendation;
• identify strengths and weaknesses of state-of-the-art multimedia search and recommendation functionalities;
• identify challenges belonging to the development of multimedia search and recommendation functionalities;
• identify evaluation criteria for multimedia search engines and recommender systems;
• explain the difference between topical relevance and utility in multimedia search and recommendation.

In addition to the core goals, students choosing the MMSR Analytics specialization will be able to:
• describe and implement cross-disciplinary approaches to multimedia search and recommendation;
• propose and justify a vision on near-future improvement opportunities for a selected state-of-the-art multimedia search and/or recommendation analytics technique.

In addition to the core goals, students choosing the MMSR Systems specialization will be able to:
• describe and implement practical solutions to deal with real-world multimedia search and/or recommendation;
• develop a practical implementation based on an academic description of a selected state-of-the-art multimedia search and/or recommendation technique and assess it against a baseline on a real-world dataset.
Education Method
lectures, lab course, specialization research or development assignment
Literature and Study Materials
Will be handed out by lecturers during the course
Assessment
Written exams (30% + 30%):
• Written partial exam over MMSR core topics (week 3, 30%);
• Written partial exam over chosen MMSR specialization (week 10/11, 30%).
For the resit, each MMSR specialization will offer one resit exam, covering the material of the two partial exams described above (60%).

Specialization assignment for chosen MMSR specialization (week 10/11, 40%):
• For MMSR Analytics: research proposal on an emerging topic in MMSR;
• For MMSR Systems: implementation of a state-of-the art MMSR research paper.

Depending on the class size, the specialization assignment may be conducted in groups. In principle, a group grade will be given to the corresponding work, unless the teaching staff sees clear motivations for differentiation in grading.

Lab assignments: pass/fail.

The following conditions on admission and grade validity apply:
• A final grade will only be constituted if a 'pass' is obtained for all lab assignments;
• A student can only join the specialization assignment and specialization exam after participation in the MMSR core exam;
• Partial results towards a final course grade (lab, partial exams, specialization assignment) do not carry over to subsequent academic years.
Special Information
Please see the Brightspace pages of this course for further information about course organization and suggested prerequisite knowledge.
Judgement
Partial exam on MMSR core (30%)
Partial exam on MMSR specialization (30%)
(Or resit on MMSR core & specialization in one exam (60%))

Specialization assignment (40%)