2D5334, Tidiga modeller och historik inom neuronnätsområdet, 2-3 poäng


The purpose of this course is to cover key ideas and models that have been central in inspiring and forming the field of neural network research. In principle, only material older than from the mid 1980ies is included. Naturally, an overview of this kind will never be complete. Here, the aim is to convey an insight into the background of contemporary concepts and to give an understanding of the process leading up to present algorithms and models.

It should be pointed out that this is not an introductory course; a level of familiarity with basic concepts and general methods at a level at least comparable to that given by e.g. course 2D1430 is required. It should also be noted that an active participation is necessary. Students are for example encouraged to search for additional material and extend the list of suggested references.

Format

Each student is responsible for one of the weekly two-hour lectures. This includes the collection and distribution of relevant material at least one week in advance as well as a presentation lasting between one hour and one hour and a half. Again, please note that the listed material is to be regarded only as a suggestion. The person responsible is strongly encouraged to add related material at discretion. Each student is also expected to serve at one occasion as "critic", i.e. to present prepared criticism and comments on the subject at hand and to start the following half-hour to one hour discussion. Attendance at at least nine occations is also required to earn the two credit points. Due to practical considerations, multiple presentations may be required in certain cases. This is then compensated for with an extra credit point, giving in total three points for the course in these cases.

Suggested material

The material below can be found in either of

Suggested lectures

  1. Turing: "On Computable numbers..."
    McCulloch & Pitts: "A Logical Calculus of the Ideas..."
    von Neumann: "Probabilistic Logic..." & "The Computer and..."
  2. Lashley: "In Search of the Engram"
    Marr: "A Theory of Cerebellar..." & "Vision"
    Hebb: "The Organization of Behavior"
    Barlow: "Single Units and Sensation..."
  3. Rosenblatt: "The Perceptron" & "A Comparison of Several..."
    Block: "The Perceptron"
    Minsky & Papert: "Perceptrons"
    Selfridge: "Pandemonium..."
    Kanerva: "Sparse Distributed Memory"
  4. Widrow & Hoff: "Adaptive Switching Circuits"
    Rumelhart et al.: "Learning Representations by Back-Propagating..."
  5. Barto & Sutton: "Associative search network..."
    (Also reinforcement in general)
  6. Uttley: "Properties of Plastic Networks"
    Steinbuch & Schmitt: "Adaptive systems using learning matrices"
    Rochester et al.: "Tests on a Cell Assembly Theory..."
    Farley & Clark: "Activity in Networks of Neuron-Like Elements"
  7. MacKay & McCulloch: "The Limiting Information Capacity of a Neuronal..."
    Eckhorn et al.: "Efficiency of Different Neuronal Codes..."
    Barlow: "The Information Capacity..."
  8. Anderson: "A Memory Storage Model..."
    Kohonen: "Associative Recall of Images"
    Amari: "Neural theory of association and concept..."
    Anderson et al.: "Distinctive features, categorical perception..."
    Willshaw et al.: "Non-Holographic Associative Memory"
  9. Hopfield: "Neural Networks an Physical..." (Discrete & continuous)
    Cragg & Temperly: "The Organization of Neurones"
    Little: "The Existence of Persistent states..."
  10. Ackley et al.: "A learning algorithm for Boltzmann machines"
    Kirkpatrick et al.: "Optimization by simulated annealing"
  11. Grossberg: Various ART models.
    Kohonen: "Self-Organized formation of topologically..."
    Willshaw et al.: "How patterned neural connections can be..."
    Nass & Cooper: "A Theory for the Development of Feature..."
    Fukushima et al.: "Neocognitron: a neural network model for..."
  12. Edelman: "Group selection As the Basis for Higher Brain..."
    Freeman: "A Physiological Hypothesis of Perception"
    Crick: "Do Dendritic Spines twitch?"
    Marr: "A Theory of Cerebellar Cortex of the Frog"

Aktuell/nästa kursomgång: period 1-2 96/97


Kursledare: Anders Lansner
Datorpostadress(er): ala@nada.kth.se

Kontakta kursledaren om du vill ha mer information om kursen.

^ Upp till kursöversikt.


Senast ändrad 1996-04-04 <blevin@nada.kth.se>