|
Center
for Adaptive Systems |
| Additional CNS Links |
| CNS brochure |
| Related
courses in other departments |
CNS Department Course OfferingsThe courses offered by the CNS Department are described below. CNS students also take a wide variety of courses in related departments. In addition, students participate in a weekly colloquium series, an informal lecture series, and a student-run Journal Club, and attend lectures and meetings throughout the Boston area; and advanced students work in small research groups.
CAS CN500: Computational Methods in Cognitive and Neural Systems
Prereq: One year of calculus or consent of instructor.This course introduces students to computer and mathematical techniques spanning a variety of scientific areas that make use of theoretical and applied computational modeling, such as engineering, mathematics, computer science and computational neuroscience. Each topic is introduced through practical examples from the literature, combining theory and applications. Topics include basic and advanced computer skills, difference and differential equations, mathematical simulation techniques, statistics, digital signal processing, control theory and image processing. The course is designed with the flexibility required to account for the varied background of participating students. 4 cr., Not offered in 2010-2011.
CAS CN510: Principles and Methods of Cognitive and Neural Modeling I
Prereq: One year of calculus and consent of instructor.Neural modeling is an interdisciplinary paradigm for discovering the computational designs that underlie human and animal learning and performance. This graduate-level course explores elements of the psychological, biological, mathematical, and computational foundations of behavioral and brain modeling. The course integrates experimental data and theoretical concepts in an interdisciplinary format. Mutually supportive constraints derived from several types and levels of analysis are used to discover organizational principles, mechanisms, local circuits, and system architectures that would otherwise be insufficiently constrained. The course presents a self-contained summary of relevant data to motivate and test key modeling ideas. Emphasis is given to analysis of the interactive, or emergent, functional properties generated by neural networks, since these properties control the behavioral success or failure of biological organisms in complex and unpredictable environments. The course presents a systematic introduction to basic mathematical principles, equations, and methods that provide a foundation for analyzing such emergent properties in key examples; notably, cooperative and competitive nonlinear feedback systems, associative learning systems, and self-organizing, self-stabilizing code-compression systems. Adaptive resonance theory is drawn upon for illustrative material because it unifies many of these themes and explains how a real-time cognitive system built from neural constituents can induce stable categories, which are fundamental for intelligent function by any cognitive system. Gorchetchnikov, 4 cr., 1st semester.
CAS CN520: Principles and Methods of Cognitive and Neural Modeling II
Prereq: One semester of linear algebra and consent of instructor.This course complements CN510, and explores the psychological, biological, mathematical and computational foundation of behavioral and brain modeling. The course introduces and analyzes ideas from three main traditions in models of learning: unsupervised (self-organized) learning, supervised learning (learning with a teacher), and reinforcement learning. By studying all three traditions in a single course, the student gains a better understanding of the strengths and weaknesses of each. Architectures studied in detail include adaptive filters, backpropagation, competitive learning, self-organizing feature maps, gradient descent procedures, the Boltzmann machine, simulated annealing, the Neocognitron, and gated dipole opponent processes. The content of the course is distinct from that of CN510, and the two may be taken concurrently. 4 cr., Not offered in 2010-2011.
CAS CN530: Neural and Computational Models of Vision
Prereq: CN510 or consent of instructor.The course acquaints advanced undergraduates and beginning graduate students with inter-disciplinary approaches to computational and neural network modeling of the functional, real-time processes of early primate vision. Topics include boundary detection, completion, and sharpening; textural segmentation and grouping; shape-from-texture and shape-from-shading; stereopsis; and motion analysis. For each process, key behavioral and physiological data will be analyzed from the standpoint of how the data constrain the computations carried out in network models of that process. Competing approaches to formal modeling will be discussed and students will carry out simulations of one or more such models on laboratory computer systems. Yazdanbakhsh, 4 cr., 2nd semester.
CAS CN540: Neural and Computational Models of Adaptive Movement Planning and Control
Prereq: CN510 or consent of instructor.This course provides an integrative treatment of a large interdisciplinary database on sensory-motor planning and control in humans and other animals. In each segment, a behavioral competence, such as the ability to maintain a stable posture, or the ability to reach to a desired target, is carefully described. Then relevant parametric data from behavioral and neurophysio-logical experiments are studied, and quantitative theoretical models are compared on the basis of their ability to explain the basic competence as well as the associated parametric database. Special emphasis is placed on models of adaptive neural networks and thereby on the process of skill acquisition. Bullock, 4 cr., 2nd semester.
CAS CN550: Neural and Computational Models of Recognition, Memory and Attention
Prereq: CN510 or consent of instructor.This course develops neural network models of how internal representations of sensory events and cognitive hypotheses are learned and remembered, and of how such representations enable recognition and recall of these events to occur. Various neural and statistical pattern recognition models are analyzed. Special attention is given to stable self-organization of pattern recognition and recall codes by Adaptive Resonance Theory (ART) models. Mathematical techniques and definitions to support fluent access to the neural network and pattern recognition literature are developed throughout the course. Experimental data and theoretical predictions from cognitive psychology, neuropsychology, and neurophysiology of normal and abnormal individuals are also analyzed. Course work emphasizes skill development, including writing, computational analysis, teamwork, and verbal communication. Ames/Gorchetchnikov, 4 cr., 2nd semester.
CAS CN560: Neural and Computational Models of Speech Perception and Production
Prereq: Consent of instructor.This course surveys aspects of anatomy, physiology, and psychophysics important for modeling hearing and speech perception. The course follows the auditory pathway from external ear to cortex, introducing relevant research areas along the way. Intended as an introductory course for students interested in pursuing research in audition and/or speech perception, topics to be covered include masking, loudness, binaural processing, auditory localization, speech perception, and models of these perceptual processes. No prerequisite courses are required; however, the course is geared towards motivated graduate students with strong quantitative skills. Some rudimentary signal processing, probability, statistics, and decision theory will be introduced in order to allow students to understand the basic material to be covered. Shinn-Cunningham, 4 cr., 1st semester (meets with ENG BE509).
CAS CN570: Neural and Computational Models of Conditioning, Reinforcement, Motivation and Rhythm
Prereq: CN510 or consent of instructor.This course develops neural and computational models of how humans and animals learn to successfully predict environmental events and generate behavioral actions which satisfy internally defined criteria of success or failure. Reinforcement learning and its homeostatic (drive, arousal, rhythm) and non-homeostatic (reinforcement) modulators are analyzed in depth. Recognition learning and recall learning networks are joined to the reinforcement learning network to analyze how these several processes cooperate to generate successful goal-oriented behavior. Maladaptive behaviors and certain mental disorders are analyzed from a unified theoretical perspective. Applications to the design of freely moving adaptive robots are noted. Tan, 4 cr., 2nd semester.
CAS CN580: Introduction to Computational Neuroscience
Prereq: Senior standing in a Natural Science or Mathematics Department or consent of instructor.This introductory level course focuses on building a background in neuroscience, but with emphasis on computational approaches. Topics include basic biophysics of ion channels, Hodgkin-Huxley theory, use of simulators such as NEURON and GENESIS, recent applications of the compartmental modeling technique, and a survey of neuronal architectures of the retina, cerebellum, basal ganglia and neo-cortex. Schwartz, 4 cr., 1st semester.
GRS CN700: Computational and Mathematical Methods in Neural Modeling
Prereq: CN500 or consent of instructor.This course introduces students to advanced techniques in computational and neural modeling. The techniques span a variety of disciplines including computer engineering, computational neuroscience, neural networks, statistics, applied mathematics, engineering, and physics. Topics such as use of simulation packages, numerical methods, statistics, control theory, differential equations, signal processing, statistical pattern recognition and vector quantization are treated on a more advanced level than in CN500. Where possible, this course has a tripartite organization. First, the theory is presented from a text or journal article. Second, students read and critique a paper that uses the technique. Finally, simulations and/or problem sets are assigned to fix the knowledge learned in the course. Pertinent examples will be drawn from research done by students and faculty in the CNS Department. Cohen, 4 cr., 2nd semester.
GRS CN710: Advanced Topics in Neural Modeling: Comparative analysis of learning systems
Prereq: CN550 or consent of instructor.This course considers the systematic analysis of supervised learning systems from neural networks, statistics, and artificial intelligence. Supervised learning systems include multi-layer perceptrons (MLP), ARTMAP, decision trees, and support vector machines. Working collaboratively, class members analyze many different algorithms and methods for pre- and post-processing data, and develop common benchmark problems and system evaluation criteria. Additional course information can be found at
http://cns.bu.edu/~gsc/CN710/pmwiki.php?n=Main.HomePage. Carpenter, 4 cr., Not offered in 2010-2011.GRS CN720: Neural and Computational Models of Planning and Temporal Structure in Behavior
Prereq: CN510 or consent of instructor; CN540 is recommended.Much of human activity consists of the formulation and execution of novel serial action plans. Serial plans are evident in all simple episodes involving preparatory actions undertaken to create the necessary conditions for a successful primary action, as well as in more complex episodes such as systematic search, communicative speech and gesture, handwriting, tool use, and object assembly. This course examines primary research literature from several relevant disciplines to identify replicable operating characteristics of serial plan formulation, choice, performance, and learning in human children and adults, with a focus on composition of novel serial plans that satisfy multiple constraints. It critically examines proposed principles governing these processes, as well as neural network (and when informative, other computationally-explicit) models that embody such principles. Bullock, 4 cr., Not offered in 2010-2011.
GRS CN730: Models of Visual Perception
Prereq: CN530 and consent of instructor.This course offers an advanced survey of selected topics of current interest in the neural and computational modeling of psychophysical and physiological data in mammalian vision. Examples of topics include visual object recognition, feature integration, computational maps, nonclassical receptive field characteristics, brightness perception, shape-from-shading, stereoscopic vision, motion perception, and optic flow. Students are expected to have a sufficient interdisciplinary grounding in the fundamentals of mammalian vision to read primary research sources extensively, and will be required to present short oral critiques of selected readings to the class. A term project that combines a literature review with formal or simulation analyses is also required. Mingolla /Yazdanbakhsh, 4 cr., 2nd semester.
GRS CN740: Topics in Sensory-Motor Control
Prereq: CN540 or consent of instructor.This course covers three main topic areas: spatial representation, speech production, and rhythmic movement. Representations appropriate for handwriting, reaching, speaking, and walking are investigated with emphasis on different levels of representation and interactions between these levels. The course covers material from psychophysics, neuroanatomy, neurophysiology, and neural modeling. Guenther, 4 cr., Not offered in 2010-2011.
GRS CN760: Topics in Speech Perception and Recognition
Prereq: CN560 or consent of instructor.This course surveys advanced topics in automatic speech recognition and auditory representation of speech signals, especially as they relate to speech perception. The course is constructed around a thorough introduction to state-of-the-art techniques in automatic speech recognition. These techniques are also related to perspectives obtained from perceptual and neurophysiological research. The course begins with the necessary fundamentals in digital signal processing and statistical pattern recognition. These are followed by detailed discussion of the major techniques in automatic speech recognition, including neural networks, hidden Markov models, and dynamic programming. The relation of these techniques to neurophysiological processing and psycholinguistic data are explored. Neural models of auditory processing and speech perception are presented and evaluated. Modeling techniques, including parameter optimization and goodness-of-fit tests, are covered. Cohen, 4 cr., 1st semester.
GRS CN780: Topics in Computational Neuroscience
Prereq: CAS MA225 Multivariate Calculus and MA242 Linear Algebra or consent of instructor.In this seminar, recent research papers and applications in computational neuroscience will be reviewed. Topics covered include cortical modeling, analog VLSI, active perception, robotic control, stereovision, and computer aided neuroanatomy. Schwartz, 4 cr., 2nd semester.
GRS CN810 Topics in Cognitive and Neural Systems: Adaptive Mobile Robotics
Prereq: CN530 or consent of instructor.Students adapt computational models for the iRobot Create to perform functions such as learning to approach or avoid objects. A term project, executed by small groups, is required, including a problem statement and an implementation of a behavioral task. Versace, 4 cr., 1st semester.
Research in Cognitive and Neural Systems
The variable-credit research courses listed below are normally open only to advanced PhD students and to students engaged in faculty-supervised research. These 900-level courses may not be used to fulfill minimum course requirements for a CNS degree.
GRS CN911, 912: Research in Neural Networks for Adaptive Pattern Recognition
GRS CN915, 916: Research in Neural Networks for Vision and Image Processing
GRS CN921, 922: Research in Neural Networks for Speech and Language Processing
GRS CN925, 926: Research in Neural Networks for Adaptive Sensory-Motor Planning and Control
GRS CN931, 932: Research in Neural Networks for Conditioning and Reinforcement Learning
GRS CN935, 936: Research in Neural Networks for Cognitive Information Processing
GRS CN941, 942: Research in Nonlinear Dynamics of Neural Networks
GRS CN945, 946: Research in Technological Applications of Neural Networks
GRS CN951, 952: Research in Hardware Implementations of Neural Networks