89-7002 Lernen und Wahrnehmen
Lernziele/Goals
- Fähigkeit, lineare und nichtlineare Filter auszuwählen, zu implementieren und anzuwenden
- Entwicklung von Mustererkennungssystemen in numerischen Umgebungen
- Fähigkeit, Daten zu analysieren und Features zu extrahieren
- Anwendung und Validierung von maschinellen Lernverfahren auf realen Daten
- Verständnis der mathematischen Grundlagen von Lernverfahren und Signalverarbeitung
Inhalte/Contents
- Arraysprachen und Programmierung für Mustererkennung
- Bayesian Methods
- Maximum Likelihood Methods
- Nearest Neighbor Classification
- Feature Extraction
- Kernel Methods
- Perceptrons
- Clustering
- Vector Quantization, k-means
- Lineare und Nichtlineare Filter
- Template Matching
- DTW, HMMs
Anwendungen in Bildverarbeitung, digitaler Photographie, multimedialen Datenbanken, Spracherkennung und Mustererkennung in Datensätzen und Textdatenbanken werden betrachtet.
http://lw.iupr.com/
89-7211 Grundlagen der Mustererkennung (Ba-Seminar)
Lernziele/Goals
- Fähigkeit zur Einarbeitung in ein spezielles Thema aus dem Bereich der
Mustererkennung
- Fähigkeit zur verstündlichen Aufbereitung, schriftlichen Beschreibung
und Präsentation eine abgegrenzten Fachthemas
- Fähigkeit zur fachlichen Diskussion
Inhalte/Contents Ausgewählte Themen der Mustererkennung, z.B.
- Klassifikation/Regression
- Lineare Klassifikatoren
- Nearest-Neighbor Klassifikation
- Logistic Regression
- Entscheidungsbäume
- Supportvektormaschinen
- Neuronale Netze
Textbook themenabhängige Literatur
89-7231 Neural Systems and Self-Organization
Lernziele/Goals
- an understanding of the formal computational and mathematical foundations of decision making
- mathematical models of neural computation and their relationship to learning algorithms
- an ability to train and apply modern parallel machine learning algorithms
- computational, statistical, neurophysiological, and psychophysical results on visual object recognition
- an understanding of selected additional research topics of current interest, including massively parallel classifier implementations, game theory, Bayesian networks
Inhalte/Contents
- review of array languages and data parallel computation
- Bayesian methods, decision theory, ML and Bayesian parameter estimation
- gradient descent methods, functions approximations, posteriors, and classification
- unsupervised learning: k-means, SOM, ICA, PCA
- McCullogh Pitts neurons, perceptrons and MLPs
- neurons and action potentials, mathematical neuron models
- computational capabilities of linear threshold units and spiking neural networks; Turing-equivalence
- neural systems as nonlinear dynamical systems
- structure and properties of the human visual system
- development and self-organization in the visual system
- feature hierarchies, HMAX, convolutional neural networks
- psychophysical results on object recognition, detection theory
- attention, salience, grouping
- additional topics of current research interest
http://ncso.iupr.com/
89-7245 Introduction to Pattern Recognition and Image Understanding (Projekt)
Das Projekt vertieft die Inhalte der Vorlesung 89-7002.
Lernziele/Goals Praktische Lösung von Mustererkennungsproblemen.
Inhalte/Contents Mögliche Projekte gibt es auf mehreren Gebieten, einschließlich Stereo- und 3D Erfassung von Objekten, Lernen von Objekterkennungsmethoden, forensische und Sicherheitsanwendungen von maschinellem Lernen und Mustererkennung, Parallisierung, Bioinformatikanwendungen, statistische Sprachverarbeitung und Text Mining.
Textbook Depending on Project
89-7253 Introduction to Image Processing and Image Understanding
This course has been replaced by 89-7002 and 89-7231. May be offered again later as a more advanced course in image processing.
Lernziele/Goals Understanding and being able to apply linear and non-linear image processing, object detection and localization, computer vision, 3D reconstruction, image matching.
Inhalte/Contents
- Image sensors
- color theory
- linear filters
- non-linear filters
- mathematical morphology
- spectral methods
- template matching
- Hough transform
- RANSAC
- digital topology
- feature extraction
- image matching.
Applications to digital photography, web imaging, object recognition, image databases, OCR, forensics, and robotics.
Textbook E.g. Gonzales and Woods: "Digital Image Processing"
http://ipiu.iupr.com
89-7254 Document and Content Analysis
Lernziele/Goals An in-depth understanding of document storage, processing, analysis, and
retrieval techniques and their applications.
Inhalte/Contents Most of the data we interact with day-to-day does not come in the form of
data structures or databases, but instead in the form of documents and
document images. This course introduces students to the formats, techniques,
and algorithms used for representing, compressing, analyzing, processing,
and displaying documents. Topics covered include:
- document formats and standards (TIFF, JPEG, PDF, PostScript, SVG)
- document image compression (G4, MRC, token based compression, JPEG2000)
- logical markup (HTML, XML, word processing formats, DocBook)
- writings systems of the world
- character sets and character encodings (ASCII, Unicode, special coding
systems)
- text rendering, layout, ligatures, and hyphenation (Pango)
- typesetting and page layout systems (text flow, Word, LaTeX, etc.)
- OCR (character recognition, page segmentation)
- spelling and orthographic variation, statistical language modeling
- document capture, page image dewarping and handheld document capture
- named entity recognition, information extraction, table recognition
- document search and retrieval, text mining, document databases
- reading, psychophysics, and human-document interaction
- document security and forensics
http://dca.iupr.com/
Textbook Will be announced in the lecture.
http://dca.iupr.com/
89-7255 Privacy, Identity and Computational Forensics
Lernziele/Goals An in-depth understanding of on-line privacy and identity, how they can be
protected, and how they can be recovered. An understanding of how
computation can be used to perform both on-line and physical forensics.
Inhalte/Contents The course covers methods for computational analysis of digital data and
physical evidence. We will be examining how personal information and
identity can be leaked and what techniques there are for protecting personal
information. Possible topics include:
- interests groups: businesses, police, intelligence agencies, criminals
- kinds of privacy and identity data
- writer identification from text
- handwriting identification from writing samples
- printer identification from printouts
- digital camera identification from images
- social network analysis
- pornography identification
- forensic applications of image data
- CAPTCHAs and reverse Turing tests
- keystroke analysis for identity
- biometric identification
- cryptography, privacy, and identity
- psychology and technology of phishing
- multi-factor authentication
- hard disk forensics
- chain of custody and standards of evidence in forensics
http://compfor.iupr.com/
Textbook Will be announced in the lecture.
http://picf.iupr.com/
89-7256 Multimedia Information Retrieval
Lernziele/Goals Multimedia information retrieval is increasingly used in online services,
such as Picasa, Google Books and Google Scholar, YouTube, and Facebook.
Students have an understanding of information retrieval systems for text,
images, videos, and audio, and the underlying statistical and algorithmic
methods.
Inhalte/Contents
- models of information retrieval
- nearest neighbors, range queries, hash algorithms
- PLSA and topic models vector space models
- text categorization, analysis, tagging, parsing
- content based image and video retrieval
- image retrieval based on color, texture, and shape
- visual bag of words model
- grouping, static and motion segmentation, scene cuts
- geometric indexing, verification, and object recognition
- automatic annotation and categorization
- HMM-based methods (audio, video, music, document retrieval)
- generative and discriminative methods
- selected special purpose applications, such as face detection, x-ray image analysis
- performance evaluation and competitions
- applications in consumer imaging, security, forensics, and copyright and plagiarism detection
Textbook
- Nilsson, N.: Artificial Intelligence: A New Synthesis, Morgan Kaufmann Publ., 1998
- Russel, S., Norvig, P.: Artificial Intelligence - A Modern Approach, Prentice Hall, 1995
- Heinsohn, J. und Socher-Ambrosius, R.: Wissensverarbeitung, Spektrum Akademischer Verlag, 1999
http://mmir.iupr.com/
89-7257 Foundations and Frontiers of Artificial Intelligence
Lernziele/Goals The course will enable students to understand and evaluate major philosophical and technological discussions surrounding AI and its applications.
Inhalte/Contents Planned topics include:
- the possibility of AI, strong/weak AI, Turing test
- philosophical objections, Chinese rooms, symbol grounding, physicality
- scientific method, automated scientific discovery
- evolution of autonomous behavior, evolutionary psychology
- synthetic biology, organic computing
- foundations from neurobiology
- Turing-Church thesis, physical limits of computation, hypercomputation and AI
- theories of consciousness
- free will, dualism
- intelligent agents and game theory
- direct brain machine interfaces, augmented cognition, mind reading
- nanotechnology, transhumanism, the singularity
- simulation hypothesis, virtual worlds
- quantum mechanics and observation
- social and intelligent agents, interaction, "laws of robotics"
- history and literature of AI and artificial agents
Textbook Will be announced in the lecture.
89-7258 Simulation and Intelligent Agents
Lernziele/Goals An understanding of the mathematical principles and models of self-organization and their computational realization and simulation. Applications in biology, economics, social sciences, and physics.
Inhalte/Contents
- mathematical models of competition, evolution, and economic activities
- Ontogeny and phylogeny
- Pattern formation and communication in biological systems.
- Multi-agent and intelligent agent modeling and analysis.
- Techniques for large-scale simulation.
- Applications of self-organization to the creation of structures.
- Applications of computational science to testing biological, social, and economic theories.
- Organic computing.
Textbook Will be announced in the lecture.
89-7259 Advanced Machine Learning
Lernziele/Goals Understanding different concepts deriving classification, modeling, and prediction for many applied domains, research level exploration of current issues and trends in the field.
Inhalte/Contents
- Support Vector Machines
- Graphical Models, Markov Random Fields
- Markov Chain Monte Carlo
- Decision Trees (CART)
- VC Dimension
- Hidden Markov Models
- exploration of current research questions in the field
Textbook “Pattern Recognition and Machine Learning” by Christopher Bishop
89-7271 Topics in Pattern Recognition (Seminar)
Lernziele/Goals
- ability to work with the original literature
- understanding of cutting-edge problems in pattern recognition and its applications
Inhalte/Contents Selected topics from pattern recognition and its applications:
- speech and handwriting recognition
- document analysis
- visual object recognition
- applications of pattern recognition to industrial problems
- bioinformatics
- learning in games
- adaptive optimization
- statistical approachs to fault tolerance in software systems
- Bayesian methods
- intelligent user interfaces
- intrusion detection
Textbook original readings
89-7281 Pattern Recognition, Machine Learning, Image Understanding (Projekt)
Lernziele/Goals Ability to design, implement, test, and validate pattern recognition and
image understanding systems.
Inhalte/Contents Available projects depend on both the research interests of the student and
research projects in the group. Areas of current interest are:
- 3D scene and object capture using stereo vision
- machine learning for image and video databases
- applying machine learning techniques to X-ray image analysis for airport security
- developing algorithms for very large machine learning problems, with applications to digital libraries and computer vision
- parallel computation for vision and pattern recognition applications (including CUDA, multicore, and distributed approaches)
Textbook Depending on Project
89-7282 Intelligent Data Mining and Pattern Recognition Competition (Projekt)
Lernziele/Goals Ability to design, implement, test, and validate pattern recognition and
image understanding systems.
Inhalte/Contents The project is intended for students interested in participating in
competitions in areas such as image retrieval, video retrieval,
recommendation systems, and classification. Every year, we pick one of these
competitions and put together a team that participates for our university.
Textbook Depending on Project.
|