|
|
 |
|
The Recognition and Categorization Group studies how the human visual system can solve a problem which no computer can yet do: recognize and categorize faces, objects, or scenes under varying viewing and illumination conditions. The problem is illustrated in the Figure: How many chairs can you find in the picture? |
|
|
|
|
|
This project studies the functional organization of auditory cortex, and which parts help to stabilize perception in noisy environments. The interactions with cognitive processes (such as attention influencing perception) are also studied. |
|
|
|
|
|
Statistical learning theory studies the process of inferring regularities from empirical data. The fundamental problem is what is called generalization: how it is possible to infer a law which will be valid for an infinite number of future observations, given only a finite amount of data? This problem hinges upon fundamental issues of statistics and science in general, such as the problems of complexity of explanations, a priori knowledge, and representation of data. |
|
|
|
|
|
Monitoring the dynamic changes of neurotransmitters (neuromodulators)in brain by MS- based approach. |
|
|
|
|
|
The Spatial Cognition Group investigates what kind of sensory information humans use for navigation and spatial orientation. For these studies virtual reality technology are used and further developed. This enables to analyze human behavior in a closed perception-action loop with naturalistic virtual environments.
 
|
|
|
|
|
|
|
|
|
|
|
|
For the study of perception an experimental paradigm was developed and elaborated for training monkeys to report what they perceive when viewing perceptually ambiguous stimuli. Such stimuli are invaluable tools for the study of the neural basis of perception and visual awareness, for they permit the dissociation of the neural responses that underlie what we perceive at any given time from those forming the sensory representation of a visual pattern. |
|
|
|
|
|
The technical equipment used to conduct our experiments ...
 |
|
|
|
|
|
The automatic acquisition and refinement of motor skills for robotics using modern machine learning techniques. |
|
|
|
|
|
Successful interactions in our environment require accurate recognition of objects. How does the brain extract and represent information about objects?
|
|
|
|
|
|
The Bayesian framework forms a consistent basis for learning, inference and decision making in the face of uncertainty. Although Bayesian inference is based on simple rules of probability theory, the solution
of practical real world problems in machine learning may require the use of sophisticated techniques including variational approximations and Markov Chain Monte Carlo. The goal of our work is to develop and assess algorithms for Bayesian learning.
|
|
|
|
|
|
The combination of functional Magnetic Resonance Imaging (fMRI) and Transcranial Magnetic Stimulation (TMS) allows us to study human brain processes with high spatial and temporal resolution. In addition, TMS can be used to demonstrate a causal relationship between the neural activity in a brain area-of-interest and behavioral performance. Currently, we use the combination of fMRI and TMS in an “offline” fashion, i.e. we first localize a brain area-of-interest using fMRI and then delineate the role of this area in greater detail using TMS. In this way, we investigate human visual perception as well as the online control of movement.
Furthermore, we are currently developing an optimized setup for interleaved TMS-fMRI, which will allow us to apply TMS while the subject is in the MR scanner. This setup will enable us to directly study the effects of TMS on human brain activity by means of fMRI.
|
|
|
|
|
|
Since its introduction in 1992, functional magnetic resonance imaging (fMRI) has grown explosively to become a standard and indispensable tool in neuroscience research in mapping functional activity in the human brain. |
|
|
|
|
|
How does the brain recognize three-dimensional objects? A first step toward understanding the neural mechanisms underlying visual object recognition can be taken by examining the nature of object representation as manifested in behavioral studies with humans or nonhuman primates. Inter-species similarities in performance allow the investigation of the neural representation of visual objects in electrophysiological experiments in monkeys. |
|
|
|
|
|
Successful interaction with our environment relies on concurrent input from different senses. Early during sensory processing, the evidences from different senses are processed separately. Yet, at some later stage, the information must be combined. This multisensory integration supposedly occurs in higher association cortices, but to some extend already in early sensory areas.
This research area investigates multisensory integration in early and higher sensory areas, especially in auditory cortex. We ask which functional auditory fields contribute to integration, how early during processing integration occurs, what anatomical connectivity mediates this integration, what role individual neurons play and what their multisensory computation is.
Especial emphasis is put on species specific communication signals. The species-specific vocalizations of non-human primates are crucial for their social interactions, reproductive success and survival. Investigating the perception and social use of vocalizations in extant non-human primates may be the most direct route to understanding the substrates underlying the evolution of speech and language. |
|
|
|
|
|
Since its early development in the late 40\'s Nuclear Magnetic Resonance (NMR) has become a powerful analytical tool for the investigation of the atomic nucleus and its environment, lending itself to applications ranging from chemical analysis or study of structures in solids to biomedical investigations. In the early 90\'s the potential of this technique for functional brain mapping was demonstrated; a fact that caused a great deal of excitement in both basic and clinical neuroscience. It was shown that by using the appropriate pulse sequences the NMR (or simply MR) imaging technique can be actually made sensitive to local magnetic susceptibility alterations produced by changes in the concentration of deoxyhemoglobin in venous blood vessels. This blood oxygenation level dependent (BOLD) contrast mechanism was successfully implemented in awake human subjects as well as in small animals such as rats and cats. |
|
|
|
|
|
Problems of computer vision and robotics are a suitable testbed both for assaying existing learning algorithms and for understanding what types of learning are required for agents - real or synthetic - to exhibit adaptive and, eventually, intelligent behaviour. |
|
|
|
|
|
How does the brain learn to make specific predictions about the environment from its sensory inputs? What are the principles that govern how the visual pathways make inferences from the visual image? How do we use image information to compute these perceptual inferences? A principal difficulty in the understanding of biological vision is the complexity of the inference problems we encounter both at the level of behavior as well as at the level of neuronal responses. This complexity mostly results from the large number of degrees of freedoms in the sensory input and in the neuronal responses. Using methods of statistical inference and learning theory, as well as signal processing, nonlinear dynamics and optimization theory, this research area addresses the problem of perceptual inference from natural images and its neural basis at different levels: (A) We develop mathematical generative models of natural images and image transformations using unsupervised learning methods. Particular emphasis is placed on quantitative comparisons of the performance of these models. (B) We perform psychophysical studies in order to evaluate the relationship between natural image models and perception. In particular, we compare perceptual predictability of different image components which are derived from different redundancy reduction models. (C) We develop new efficient methods to predict the spike trains of neurons in response to natural stimuli with the goal of inferring the contribution of these neurons to the image processing performed in the early visual system. In particular, we build population response models for multi-cell recordings and we address the aspects of contrast adaptation, non-Gaussian stimuli, and inter-spike correlations. |
|
|
|
|
|
The primary purpose of perception is to enable interaction with the environment. Interactions may include walking from one place to another, grasping an object, talking to a person, or navigating a car. In return such actions directly affect our perceptions of the world. These interdependencies between action and perception are illustrated by the “Action/Perception-Cycle” (see Figure below). The view we take in sensorimotor integration is that in many aspects of behavior, motor actions and sensory processing are inseparably linked and therefore have to be studied in a closed action/perception loop.
|
|
|
|
|
|
We are studying the principled design of learning algorithms that are able to identify regularities in data. Subjects of research in this area contain are not only the development of improved algorithms for tasks such as pattern recognition, regression estimation, density estimation, and novelty detection, but also the formalization of new learning problems, and the design of algorithms for solving these problems.
The general framework in which we tackle these issues is that of kernel learning algorithms (e.g., Support Vector machines). One can show that a certain class of kernels corresponds to dot products in so-called reproducing kernel Hilbert spaces. These kernels allow the generalization of various algorithms to nonlinear settings. |
|
|
|
|
|
The study of intelligent systems is a study of adaptive behavior; in other words, it is the study of problem solving, learning, and evolution. Perceiving, remembering, forgetting, or recognizing are also adaptive processes that can be only realized because of the brain's capacity to reorganize its regional structural and functional organization. Such plasticity also underlies the reorganization occurring during brain damage or sensory deprivation. |
|
|
|
|
|
Genomics and proteomics are currently producing massive datasets which contain a wealth of information that is going to transform medicine in the years to come. Unlike naturally occuring high-dimensional measurements such as visual images or auditory signals, measurements in bioinformatics are not easily interpretable by human observers. For a learning machine, on the other hand, there is no fundamental difference between these different types of data. In this domain, machine intelligence will likely prove indispensable to make sense out of the data. |
|
|
|
|
|
New classes of 'targeted' and 'smart' MR contrast agents are being developed to report on the physiological status or metabolic activity of biological systems. The goal is therefore to develop new cell-specific contrast agents which are either internalized or trapped selectively in the target cells for structural MRI studies or tracking of cells in vivo by MRI. |
|
|
|
|
|
In the Perception, Graphics & Computer Vision (PGCV) group, our aim is to bring computer vision, computer graphics and psychophysics together in order to investigate fundamental perceptual and cognitive processes. The fusion of methods and approaches from these three research areas has the potential to greatly advance our understanding of the processes in the human brain. |
|
|
|
|
|
Using psychophysical experiments with human observers, computer simulations and mathematical models we investigate how the human visual system processes and represents data, and how its properties are tuned to the statistics of naturally occurring stimuli. |
|
|
|
|
|
A brain computer interface is a system directly or indirectly connected to the brain, that allows its user to communicate with its environment without the use of muscles. |
|
|
|
|
|
|
