AIMS AND CONTENT

LEARNING OUTCOMES

Neuromorphic paradigms for the representation and distributed processing of multidimensional signals. Computational primitives and architectural models. Elements of cognitive science and cognitive robotics for the development of autonomous behavior in natural environments and in complex systems.

LEARNING OUTCOMES (FURTHER INFO)

Inform the student on modern integrated approaches to the study of cognitive systems that take into account (1) the principles of organization and development of the nervous system, (2) the co-existence of the nervous system in a body, and therefore their mutual conditioning (embodiment problem), (3) the complexity of the external environment, as well as (4) the interactions between different sensory modalities. In this framework, the course proposes methods, techniques and tools for analysis, simulation and synthesis of neuromorphic perceptual engines for sub-symbolic representation of sensory information, and its functional integration with motor behavior. Visuo-spatial perception, sensorimotor coordination, and control capabilities are taken as the reference framework. The approach combines principles from signal processing, connectionism, and embodied cognition.

TEACHING METHODS

Traditional lectures (48 hours) & lab experiences (8 hours).

SYLLABUS/CONTENT

-    Multichannel distributed representation of multidimensional sensory signals.
-    Receptive fields and neuromorphic operators: linear models, static (and dynamic) non-linearities, energy models.
-    Multilayer cortical-like architectures: computational primitives and interconnection schemes, neural fields, lateral inhibition, recursive networks, winner-takes-all circuits, normalization circuits and gain fields.
-    Tuning curves and population coding: uncertainty and intrinsic noise, predictive coding, probabilistic interpretation, optimal codes, and decoding techniques.
-    Perceptual engineering: computational visual perception. Motion and depth processing.
-    Elements of cognitive science and cognitive robotics for the development of autonomous behaviors in natural real-world conditions.
-    Implications of perception-action cycles on computational modeling.
-    Lab experiences on perceptual engines (features, potentialities and application domains) and on ecological perception-action cells (integrated environments to assess, monitor, influence, and exercise perceptual experiences in closed loop with action).

RECOMMENDED READING/BIBLIOGRAPHY

Slides and other distributed material (available through Aulaweb).

Recommended texts:

• P. Dayan and L.F. Abbott. Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems. The MIT Press, 2001.
• H.A. Mallot. Computational Vision: Information Processing in Perception and Visual Behavior. The MIT Press, 2000.

TEACHERS AND EXAM BOARD

Exam Board

SILVIO PAOLO SABATINI (President)

FABIO SOLARI

LESSONS

Class schedule

NEUROMORPHIC COMPUTING AND INTEGRATIVE COGNITIVE SYSTEMS

EXAMS

EXAM DESCRIPTION

50% from oral examination. The oral examination consists in properly placing and discussing two given topics (typically a 'theoretical model' and 'computational perceptual problem').

50% from the evaluation of a group project on examples and/or supplementary analysis of topics presented in classes.Evaluation will be based on: (1) comprehension of the problem, (2) appropriateness and (3) completeness of the proposed solution, (4) organization and clarity of the results.

ASSESSMENT METHODS

At the end of the course, the student should demonstrate:

• Capability of analysing and synthetizing neuromorphic computational modules at single cels, circuit, and system levels.
• Capability of designing experiments for (1) the validation of functional models in controlled natural conditions, (2) the assessement of the correct development of perceptual and sensorimotor abilities, (3) the set-up of neuro-cognitive and neuro-motor rehabilitation exercises.

FURTHER INFORMATION