Conceptual Spaces for Artificial Intelligence

6 EC

Semester 1, period 1

5082CSFA6Y

Owner Bachelor Kunstmatige Intelligentie
Coordinator Martha Lewis
Part of Minor Kunstmatige Intelligentie, year 1Minor Logic and Computation, year 1Bachelor Kunstmatige Intelligentie, year 3

Course manual 2018/2019

Course content

Conceptual spaces are a way of representing concepts geometrically that can be used in AI, psychology, and language representations. Conceptual spaces lie between neural and symbolic models and therefore form a bridge between these two levels of representation. The course covers understanding what conceptual spaces are and how they can be formalized, how they can be used in AI applications, and how they are used for language representation. The course will involve theoretical work, programming in Python or MATLAB, and discussion of difficulties and areas for research in conceptual spaces.

The field of conceptual spaces is relatively new and an area of active research with many open questions. This course would particularly suit students who would potentially be interested to apply some of their learning in a summer project.

Objectives

By the end of the course the students should:

  • Understand conceptual spaces and how concepts are represented

  • Understand some basic fuzzy set theory and why this is used in concept representations

  • Be able to represent data within a conceptual space and build concepts within the space

  • Be able to use a conceptual space representation to categorize objects and form combined concepts

  • Understand conceptual spaces for language representation including vector spaces semantics and how formal semantics can be applied.

Teaching methods

  • Lecture
  • Seminar
  • Self-study

The lectures will provide the theoretical background to the course, and lay out the techniques to be used in the assignments. Guest lecturers will provide the students with an insight into the latest research in conceptual spaces

The seminars will provide practice in using the techniques and theory described in the lectures. Teaching assistants will demonstrate techniques and be on hand to help with homework questions.

Assignments will be completed in self-study time and serve to consolidate the ideas introduced in lectures and seminars.

Learning activities

Activity

Hours

Hoorcollege

24

Tentamen

2

Tussentoets

2

Werkcollege

24

Self study

116

Total

168

(6 EC x 28 uur)

Attendance

Programme's requirements concerning attendance (OER-B):

  • For practical trainings and tutorials with assignments attendance is obligatory. The requirements for attendance might differ between courses and are stated in the course manual. When students do not meet the requirements for attendance, he or she cannot finish the course with a pass mark.

Assessment

Item and weight Details

Final grade

20%

Tussentoets

20%

Tentamen

15%

Homework 1

Programming

Calculations and discussion

15%

Homework 2

Programming

Calculations and discussion

15%

Homework 3

Programming

Calculations and discussion

15%

Homework 4

Programming

Calculations and discussion

Students must obtain a grade of 9/20 in at least one exam. The exams will be open-book: access to study materials is allowed.  Computers may be used to look at slides, practice exercises, homeworks etc, and any printed material. Either a computer or a calculator is needed to do some of the calculations. Internet and phones may not be used.

The tussentoets will be based on weeks 1-3 and the tentamen will be based on weeks 5-7. The resit will consist of two exam papers, students can choose whether to resit the tussentoets or the tentamen, but not both.

Deadlines for the homeworks are strict: failure to meet the deadline will result in a score of 0 for that homework.

Students must obtain an overall grade in line with the UvA minimal passing grade to pass the course.

Inspection of assessed work

Students will be notified by email in DataNose or Canvas.

Assignments

There are 4 homework assignments. Each of these carries 15% of the final grade and is scored out of 15. Each homework will consist of a programming element (10 points) and a calculation and discussion element (5 points). Homeworks should be completed in groups of 2-3. Homework assignments are graded and count towards the final exam. Feedback will be given by TAs in the werkcolleges.

Fraud and plagiarism

The 'Regulations governing fraud and plagiarism for UvA students' applies to this course. This will be monitored carefully. Upon suspicion of fraud or plagiarism the Examinations Board of the programme will be informed. For the 'Regulations governing fraud and plagiarism for UvA students' see: www.student.uva.nl

Course structure

Weeknummer Onderwerpen Studiestof
1

Week 1: Introduction

  • Why conceptual spaces? Categorization; similarity; route from neural to symbolic representations

  • What is a conceptual space? Quality dimensions; domains; integral and separable dimensions

  • Concept representation: Properties; Concepts; Prototype theory; Voronoi tessellations and other approaches

Lecture notes, Practice exercises, Homework 1
2

Week 2: Vagueness and Computing with Conceptual Spaces

  • Introduction to fuzzy set theory: Membership and similarity; connectives; links to prototype theory
  • Difficulties: ‘Pet fish’ problem; psychological evidence for overextension and underextension; stereotypes/extremes
  • Formalizations of conceptual spaces: Aisbett and Gibbon, Lawry and Tang, Adams and Raubal, Bechberger
  • Computing combinations of concepts: Fuzzy intersections, contrast classes
Lecture notes, Practice exercises, Homework 2
3

Week 3: Computing with Conceptual Spaces cont’d

  • Building conceptual spaces: Building a conceptual space from data, clustering points in a space, extracting meaningful dimensions from data

  • Combining concepts in conceptual spaces: examples from Schockaert/Bechberger

Lecture notes, Mock exam paper
4 Week 4: Tussentoets covering weeks 1-3 Lecture notes from previous weeks, Mock exam paper
5

Guest Lectures:

  • Peter Gärdenfors

  • Steven Schockaert

Lecture notes, Practice exercises, Homework 3
6

Week 6: Conceptual Spaces and Language

  • Introduction to formal semantics: Set-theoretic meaning, computing truth values of sentences
  • Semantic spaces: vector representations of words, relevance of space dimensions, methods of computing meaning vectors
  • Beyond nouns: Adjectives and verbs in conceptual spaces
Lecture notes, Practice exercises, Homework 4
7

Week 7: Grammar in Conceptual Spaces

  • Mapping grammar onto conceptual spaces: Pregroup grammar, adjectives and verbs in semantic spaces and conceptual spaces

  • Combining words in conceptual spaces: type reductions, examples.

  • Guest lecture: Mehrnoosh Sadrzadeh

Lecture notes, Mock exam paper
8 Week 8: Tentamen covering weeks 5-7 Lecture notes from previous weeks, Mock exam paper

Timetable

The schedule for this course is published on DataNose.

Processed course evaluations

Below you will find the adjustments in the course design in response to the course evaluations.

Contact information

Coordinator

  • Martha Lewis