PSY 5038W - Introduction to Neural Networks

Fall 2014
Class #: 34641  
9:45AM-11:00AM MW
Elliott Hall N227, TCEASTBANK

Course home pages: courses.kersten.org

Instructor: Daniel Kersten, kersten@umn.edu, Office: S212 Elliott Hall, Phone: 625-2589
Office hours: Mondays 11:00 to 12:00 and by appointment.

Teaching assistant: Cheng Qiu, qiuxx077@umn.edu
Office hours:
Tuesdays 11:00 to 12:00

Course description. Introduction to large scale parallel distributed processing models in neural and cognitive science. Topics include: linear models, statistical pattern theory, Hebbian rules, self-organization, non-linear models, information optimization, and representation of neural information. Applications to sensory processing, perception, learning, and memory.

Prerequisites: Linear algebra, multivariate calculus.

Readings

Lecture notes (see below)

Software

Mathematica

Mathematica is the primary programming environment for this course. If you wish to purchase Mathematica for Students see http://www.wolfram.com/products/student/mathforstudents/index.html).

Accessing Mathematica on the CLA servers:

For user help on using Mathematica, see: http://mathematica.stackexchange.com

Learning center: http://www.wolfram.com/learningcenter/

Python/IPython

Supplementary

Writing

Gopen, G. D., & Swan, J. A., 1990. The Science of Scientific Writing. American Scientist, 78, 550-558. (pdf)

Supplementary

Writing assistance

THE CENTER FOR WRITING offers free one-to-one writing assistance to undergraduate and graduate students, with appointments up to 45 minutes.
Nonnative speaker specialists are available. For more information, see http://writing.umn.edu

Supplementary readings

*Anderson, James. (1995) Introduction to Neural Networks, MIT Press.
***Bishop, C. M. (2006). Pattern recognition and machine learning. New York: Springer.
*Dayan, P., & Abbott, L. F. (2001). Theoretical neuroscience : computational and mathematical modeling of neural systems. Cambridge, Mass.: MIT Press.
Freeman, J. A. (1994). Simulating Neural Networks with Mathematica . Reading, MA: Addison-Wesley Publishing Company. http://library.wolfram.com/infocenter/Books/3485/
**Gershenfeld, N. A. (1999).
The nature of mathematical modeling. Cambridge ; New York: Cambridge University Press.
**Hertz, J., Krogh, A., &;Palmer, R. G. (1991). Introduction to the theory of neural computation (Santa Fe Institute Studies in the Sciences of Complexity ed.). Reading, MA: Addison-Wesley Publishing Company.
Koch, C., & Segev, I. (Eds.). (1998). Methods in Neuronal Modeling : From Ions to Networks (2nd ed.). Cambridge, MA: MIT Press.
***MacKay, D. J. C. (2003). Information theory, inference, and learning algorithms. Cambridge, UK ; New York: Cambridge University Press. http://www.inference.phy.cam.ac.uk/mackay/itila/book.html
***Murphy, K. P. (2012). Machine Learning: a Probabilistic Perspective. MIT Press.

*Neural/Cognitive Science
**Physics/Applied Math
***Statistical/machine learning

Grade Requirements

There will be a mid-term, final examination, programming assignments, as well as a final project. The grade weights are:


Outline & Lecture Notes
(under construction)

http://onestop.umn.edu/calendars/#fall2014

(NOTE: Links to revised lecture material below will be posted on the day of the lecture.
Links to the pdfs for additional readings may require a password.
If you want to preview similar material, check out lectures from 2009
)


Lecture notes are in Mathematica Notebook and pdf format. You can download the Mathematica notebook files below to view with Mathematica or Wolfram CDF Player (which is free).

 

Date

Lecture

Additional Readings & supplementary material

Assignments
due

1

Sep 3

Introduction
(pdf file)|Mathematica notebook

Mathematica screencast
Neuroscience tutorial (Clinical, Wash. U.)
Top 100 Brain Structures
fMRI4Newbies

 

2

Sep 8

The neuron (pdf file)| Mathematica notebook

Hodgkin-Huxley.nb
Koch & Segev, 2000 (pdf)
Meunier & Segev, 2002 (pdf)

3

Sep 10

Neural Models, McCulloch-Pitt (pdf file)| Mathematica notebook

Koch, C., & Segev, I. (Eds.). (1998) (pdf)

 

4

Sep 15

Generic neuron model (pdf file)| Mathematica notebook

   

5

Sep 17

Lateral inhibition (pdf file)| Mathematica notebook

Hartline (1972) (pdf)

 

6

Sep 22

Matrices (pdf file)| Mathematica notebook

  Homework 1. Mathematica notebook

7

Sep 24

Linear systems, learning & memory (pdf file)|
Mathematica notebook

   

8

Sep 29

Linear association and memory simulations (pdf file)| Mathematica notebook

einstein64x64.jpg
shannon64x64.jpg

 

9

Oct 1

Non-linear networks, discriminative models, Perceptron, SVMs (pdf file)| , Mathematica notebook


Preview of statistical sampling (pdf) Mathematica notebook
LeNet-5

SVMs: Jäkel et al., (2009) Jäkel et al., (2007)
Mathematica SVMs: Nilsson, Björkegren & Tegnér (2006)
MathSVMv7.nb

Fisher's linear discriminant notes (pdf) Mathematica notebook
(updated)

 

 

10

Oct 6

Supervised learning as regression, Widrow-Hoff, backprop (pdf file)| Mathematica notebook

Backpropagation.m

XOR backpropagation demo. Mathematica notebook

Poirazi, Brannon & Mel (2003) (pdf)

Williams (1992) (pdf)

11

Oct 8

Hopfield networks (pdf file)| Mathematica notebook

Hopfield (1982) (pdf)
Marr & Poggio (1976) (pdf)
Hopfield (1984) (pdf)
Durstewitz et al. (2000) (pdf)

Ipython demo of Hopfield

Stereo correspondence. Mathematica demo.

 

12

Oct 13

Boltzmann machine (pdf file)| Mathematica notebook

Sculpting the energy function, interpolation (Mathematica notebook)

Berkes, P., Orban, G., Lengyel, M., & Fiser, J. (2011). Spontaneous cortical activity reveals hallmarks of an optimal internal model of the environment. Science, 331(6013), 83. (pdf)

Homework 2. Lateral inhibition, Widrow-Hoff, back-prop

13

Oct 15

Probability and neural networks
(pdf file)| Mathematica notebook

ProbabilityOverviewNN.nb
Griffiths and Yuille (2006) (pdf)
Jordan, M. I. and Bishop. C. MIT Artificial Intelligence Lab Memo 1562, March 1996. Neural networks.

Kersten, D., & Yuille, A. (2003). Bayesian models of object perception. Current Opinion in Neurobiology, 13(2), 1-9. (pdf)

Kersten D. & Yuille, A.L (2013) .Vision: Bayesian Inference and Beyond. The New Visual Neurosciences. John S. Werner and Leo M. Chalupa (Editors) MIT Press. Cambridge MA..(draft pdf)

 

14

Oct 20

Probability continued
(pdf)
Mathematica notebook

Pattern Recognition and Machine Learning, Chapter 8: Graphical Models. Christopher M. Bishop (pdf)

 

15

Oct 22

Regression, Interpolation, perceptual completion, bias/variance
(pdf)
Mathematica notebook

Weiss Y. (pdf)
Belief propagtion tutorial by James Coughlan (pdf).
Ma, W. J. (2012). Organizing probabilistic models of perception. Trends in Cognitive Sciences, 16(10), 511–518. (pdf)

PROJECT IDEAS (pdf)
Sample abstracts from past students

For demonstration style projects, see the Wolfram Demonstration site.

A specific example.

 

Oct 27

MID-TERM TEST

MID-TERM STUDY GUIDE

MID-TERM (16%)

16

Oct 29

Belief Propagation: regression and interpolation revisited (pdf )

Mathematica notebook

 

James Coughlan's BP tutorial

Geisler, W. S., & Kersten, D. (2002). Illusions, perception and Bayes. Nat Neurosci, 5(6), 508-510. (pdf)

 

17

Nov 3

 

 

Utility & probabiilty: Bayes decision theory (pdf)
Mathematica notebook

Neural networks in the context of machine learning

How To Do Research. William T. Freeman (2013), (link)

 

18

Nov 5

Supervised learning: neural networks in the context of machine learning cont'd

(pdf)
Mathematica notebook

 

 

 

Anaconda python installation recommended. We will use IPython, a browser-based notebook interface for python.

See here for illustrations of IPython cell types, and here for a collection of sample notebooks.

Look here for some good tips on installation, as well as the parent directory for excellent ipython-based course material on scientific computing using Monte Carlo methods.

For a quick start to scientific programming, see: http://nbviewer.ipython.org/gist/rpmuller/5920182

For a comphrensive coverage of scientific python see:https://scipy-lectures.github.io

And for a ground-up set of tutorials on python see: http://learnpythonthehardway.org/book/

Switching from matlab to python? http://wiki.scipy.org/NumPy_for_Matlab_Users

 

19

Nov 10

Overview of python/ipython for scientific compution/neural networks, and Bayesian computations.

IPython notebook
Live notebook
(pdf)


 

Practice: 1) TwoNeuroHopfield.py, 2) convolution, which needs Zebra_running_Ngorongoro.jpg

 

20

Nov 12

More sampling
MRFs
Gibbs sampling
MCMC

Mathematica notebook

(pdf)

Metropolis2D.ipynb

http://pymc-devs.github.io/pymc/

Recommended pymc tutorials: http://camdavidsonpilon.github.io/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers/

Homework 3

21

Nov 17

PyMC (IPython notebook)

Carrandini, Heeger, Movshon (1996)(pdf)

Supplement (pdf file)| Mathematica notebook
smallRetinaCortexMap.nb
GraylefteyeDan.jpg

Kersten D. & Yuille, A.L. (2014) Inferential Models of the Visual Cortical Hierarchy. The New Cognitive Neurosciences, 5th Edition.(draft pdf)


Fang, F., Boyaci, H., & Kersten, D. (2009). Border ownership selectivity in human early visual cortex and its modulation by attention. J Neurosci, 29(2), 460-465.

Final project title & paragraph outline due (2%)

22

Nov 19

Architectures: Overview of visual hierarchy: Lect_22a_VisualArchitecture(Keynote pdf)

Lect_22b_AdaptMaps(pdf)| Mathematica notebook

 

Simoncelli, E. P., & Olshausen, B. A. (2001). Natural image statistics and neural representation. Annu Rev Neurosci, 24, 1193-1216.(pdf)
Supplement: ContingentAdaptation.nb

Contrast normalization notes

Oja's rule and PCA: Sanger (2003) (pdf)



 23

Nov 24

keynote presentation (pdf)

Neural networks for self-organization, efficient coding,Principal Components

Mathematica notebook (pdf) d

 

Knill & Pouget (2004) (pdf)
Pouget et al. (2006) (pdf)
Ernst, M. O., & Banks, M. S. (2002). Humans integrate visual and haptic information in a statistically optimal fashion. Nature, 415(6870), 429-433. (pdf)

 

Quiroga, R. Q., Reddy, L., Kreiman, G., Koch, C., & Fried, I. (2005).(pdf)

24

Nov 26

Probabilistic neural representations and the neural integration of information

keynote presentation (pdf)

Mathematica notebook (pdf)

 

 

Ma, W. J. (2012). Organizing probabilistic models of perception. Trends in Cognitive Sciences, 16(10), 511–518. (pdf)

 

 


 

25

Dec 1

Scientific writing and presentations (pdf)
(Mathematica notebook)

 

Gopen & Swan, 1990 (pdf)

Denis Pelli's advice for scientific writing

(no Homework 4 )

26

Dec 3

Clustering, EM, segmentation
Mathematica notebook (pdf)

Expectation Maximization: Weiss Y. (pdf)

Kirchner, H., & Thorpe, S. J. (2006). Ultra-rapid object detection with saccadic eye movements: Visual processing speed revisited. Vision Research, 46(11), 1762–1776. doi:10.1016/j.visres.2005.10.002 (pdf)

Ullman, S., Vidal-Naquet, M., & Sali, E. (2002). Visual features of intermediate complexity and their use in classification. Nat Neurosci, 5(7), 682-687. (pdf)

Hegde, J., Bart, E., & Kersten, D. (2008). Fragment-Based Learning of Visual Object Categories. Curr Biol. 18, 597-601
(pdf)

Serre, T., Oliva, A., & Poggio, T. (2007). A feedforward architecture accounts for rapid categorization. Proc Natl Acad Sci U S A, 104(15), 6424-6429.
(pdf)

 

 

27

Dec 8

Bidirectional hierarchical models

keynote presentation (pdf)

Bullier, J. (2001). Integrated model of visual processing. Brain Res Brain Res Rev, 36(2-3), 96-107. (pdf)

Epshtein, B., Lifshitz, I., & Ullman, S. (2008). Image interpretation by a single bottom-up top-down cycle. Proceedings of the National Academy of Sciences of the United States of America, 105(38), 14298. (pdf)

Fang, F., Boyaci, H., Kersten, D., & Murray, S. O. (2008). Attention-dependent representation of a size illusion in human V1. Curr Biol, 18(21), 1707-1712 (pdf)

Yuille, A., & Kersten, D. (2006). Vision as Bayesian inference: analysis by synthesis? Trends Cogn Sci, 10(7), 301-308. (pdf)

Rao, R. P., & Ballard, D. H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nat Neurosci, 2(1), 79-87. (pdf)

Kalman notes (pdf)
Kalman tracking demo (Mathematica notebook)
Wolpert et al (1995) (pdf)

Complete Draft of Final Project (5%) Due December 8

 

 

Dec 10

Last day
of instruction

In-class Final Test

FINAL STUDY GUIDE

Peer comments on Final Project (5%:) Due Friday December 12

FINAL TEST (16%)

 

Dec 15

   

Drafts returned with Instructor comments December 15

 

Dec 17

     
 

Dec 18

End of Semester

  Submit Final Revised Draft of Project (28%)

Final Project Assignment

This course teaches you how to understand cognitive and perceptual aspects of brain processing in terms of computation. Writing a computer program encourages you to think clearly about the assumptions underlying a given theory. Getting a program to work, however, tests just one level of clear thinking. By writing about your work, you will learn to think through the broader implications of your final project, and to effectively communicate the rationale and results in words.

Your final project will involve: 1) a computer simulation and; 2) a 2000-3000 word final paper describing your simulation. For your computer project, you will do one of the following: 1) Devise a novel application for a neural network model studied in the course; 2) Write a program to simulate a model from the neural network literature ; 3) Design and program a method for solving some problem in perception, cognition or motor control. The results of your final project should be written up in the form of a short scientific paper, describing the motivation, methods, results, and interpretation. Your paper will be critiqued and returned for you to revise and resubmit in final form. You should write for an audience consisting of your class peers. You may elect to have your final paper published in the course's web-based electronic journal.

Completing the final paper involves 3 steps:

  1. Outline. You will submit a working title and paragraph outline by the deadline noted in the syllabus. These outlines will be critiqued in order to help you find an appropriate focus for your papers. (2% of grade). (Consult with the instructor or TA for ideas well ahead of time).
  2. Complete draft. You will then submit a complete draft of your paper (2000-3000 words). Papers must include the following sections: Abstract, Introduction, Methods, Results, Discussion, and Bibliography. Use citations to motivate your problem and to justify your claims. Figures should be numbered and have figure captions. Cite authors by name and date, e.g. (Marr & Poggio, 1979). Use a standard citation format, such as APA. Papers must be typed, with a page number on each page.Each paper will be reviewed with specific recommendations for improvement. (5% of grade)
  3. Peer commentary. Each student will submit a paragraph on an anonymous paired project draft (5% of grade)
  4. Final draft. You will submit a final revision for grading. (28% of grade). The final draft must be turned in by the date noted on the syllabus. Students who wish to submit their final papers to be published in the class electronic journal should turn in both paper and electronic copies of their reports.

If you choose to write your program in Mathematica, your paper and program can be combined can be formated as a Mathematica notebook. See: Books and Tutorials on Notebooks.

Your paper will be critiqued and returned for you to revise and resubmit in final form. You should write for an audience consisting of your class peers.

Some Resources:

Student Writing Support: Center for Writing, 306b Lind Hall and satellite locations (612.625.1893) http://writing.umn.edu.
Online Writing Center: http://writing.umn.edu/sws/visit/online/index.html

NOTE: Plagiarism, a form of scholastic dishonesty and a disciplinaryoffense, is described by the Regents as follows: Scholasticdishonesty means plagiarizing; cheating on assignments or examinations;engaging in unauthorized collaboration on academic work; taking,acquiring, or using test materials without faculty permission; submittingfalse or incomplete records of academic achievement; acting alone or incooperation with another to falsify records or to obtain dishonestlygrades, honors, awards, or professional endorsement; or altering,forging, or misusing a University academic record; or fabricating orfalsifying of data, research procedures, or data analysis.http://regents.umn.edu/sites/regents.umn.edu/files/policies/Student_Conduct_Code.pdf

2014 Daniel Kersten, University of Minnesota