9.641 Neural Networks

Lecture schedule

• Lecture 1 (Th 9/5) (Postscript, PDF, realvideo)
• Definitions of computational neuroscience and neural networks.  Classical neural network equations.  Integrate-and-Fire model neurons and reduction by the method of averaging.
• Christof Koch, Biophysics of Computation, Section 14.2: 335-341, Oxford University Press, New York, Oxford (1999).
• Bard Ermentrout, Reduction of Conductance-Based Models with Slow Synapses to Neural Nets, Neural Computation. 6,679-695 (1995).
• Optional Recitation 1 (M 9/9, 7pm in E25-401) (realvideo)
• Discuss homework 1. Review of basic differential equations, Taylor approximations, and MATLAB skills for homework 1.
• Lecture 2 (T 9/10) (Postscript, PDF, realvideo)
• Perceptron as feature detector. Visual receptive fields.
• D. Marr, Vision, Section 2.2, 54-79, W.H. Freeman and Company, New York (1982).
• D.H. Hubel, Eye, Brain, and Vision, Chapter 3, 39-46, Scientific American Library, New York, (1988-1995).
• Assignment 1: Integrate-and-fire neurons, method of averaging.
• Lecture 3 (Th 9/12) (Postscript, PDF, realvideo)
• The problem of credit assignment.  Perceptron learning rule.  Convergence theorem.  Learning by gradient following.  Online learning. 
• Hertz, Krogh, and Palmer Chapter 5.
• Optional Recitation 2 (F 9/13, 2-3pm in E25-401) (realvideo)
• Homework 2. Manipulating the MNIST database in MATLAB. Eigenvalues. Plotting contours in MATLAB.
• Lecture 4 (T 9/17) (Postscript, PDF,realvideo)
  • Multilayer perceptrons and backpropagation.
  • Hertz, Krogh, and Palmer Chapter 6.
  • Y. LeCun, L. Bottou, G. B. Orr, K.-R. Muller. Efficient backprop, in G. Orr and K. Muller, Neural networks: tricks of the trade, Springer, 1998.
  • Assignment 2: Perceptrons.
• Lecture 5 (Th 9/19) (realvideo)
• Backpropagation applications. LeNet and the visual system
• Y. LeCun, Y. Bengio, Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, Gradient-based learning applied to document recognition, Proc. IEEE, November 1998.
• David A. Robinson, Implications of neural networks for how we think about brain function, Behav. Brain. Sci., 15, 644-55 (1992).
• Lecture 6 (T 9/24) (Postscript, PDF,realvideo)
• The capacity of the perceptron, statistical learning theory.
• Hertz, Krogh, and Palmer Section 5.7
• T. M. Cover, Geometrical and Statistical Properties of Systems of Linear Inequalities with Applications in Pattern Recognition, IEEE Trans. Electronic. Comput. 14, 326-34 (1965).
• V. Vapnik, The nature of statistical learning theory, Springer (1995).
• Assignment 3. Backpropagation
• Lecture 7 (Th 9/26) (Postscript, PDF,realvideo)
• Unsupervised learning for perceptrons. Mean and principal component.
• Lecture 8 (T 10/1) (Postscript, PDF,realvideo)
• Feedback in linear networks.  Eigenmode analysis, amplification and attenuation, gain-bandwidth theorem.
• G. Strang, Introduction to Applied Mathematics, Section 1.5, 47-68, Wellesley-Cambridge Press, Wellesley, Massachusetts (1986).
• Lecture 9 (Th 10/3) (realvideo)
• Neural network models of the retina.
• W. H. Press et al., Numerical Recipes in C, 2d ed., Chaps. 12 and 13.
• G. Strang, Introduction to Applied Mathematics, Section 4.2, 290-309, Wellesley-Cambridge Press, Wellesley, Massachusetts (1986).
• E. H. Adelson, Lightness perception and lightness illusions.
• H. K. Hartline and F. Ratliff, "Inhibitory Interaction in the Retina of Limulus". Physiology of Photoreceptor Organs, Ed. M.G.F. Fuortes, 382-447, Springer-Verlag, Berlin, Heidelberg, New York (1972)
• C. Mead, Analog VLSI and Neural Systems, Appendix C, 339-351, Addison-Wesley Publishing Company (1989)
• E.R. Kandel, J.H. Schwartz, T.M. Jessell, Principles of Neural Science, 4/e, Part V, Chapter 26, 507-522, McGraw-Hill (2000)
• K.A. Boahen, Spatio-temporal Sensitivity of the Retina: A Physical Model, 1-19, Computation and Neural Systems Program, California Institute of Technology, Pasadena, CA (1991).
• Lecture 10 (T 10/8) (Postscript, PDF,realvideo)
• Self-excitation and global inhibition. Decision-making. The MAX operation.
• Assignment 5:  Linear network theory
• Lecture 11 (Th 10/10) (Postscript, PDF,realvideo)
• Hybrid analog-digital computation. Permitted and forbidden sets.
• R. Hahnloser, R. Sarpeshkar, M. Mahowald, R.J. Douglas, H.S. Seung, Digital selection and analog amplification coexist in a cortex-inspired silicon circuit, Nature, 405, 947-51, 2000.
• Columbus Day holiday (T 10/15)
• Midterm review (W 10/16)
• Midterm exam (Th 10/17)
• Lecture 12 (T 10/22) (realvideo)
• Intra-group excitation and global inhibition. Marr-Poggio model of stereopsis. Complex cell model.
• D. Marr, Vision, Section 3.3, 111-159, W.H. Freeman and Company, New York (1982).
• Assignment 6: Nonlinear network theory.
• Lecture 13 (Th 10/24) (realvideo)
• Lateral excitation and global inhibition.  Gain fields and stimulus selection.
• R. Ben-Yishai, R. L. Bar-Or, and H. Sompolinsky, Theory of orientation tuning in visual cortex, PNAS, 92, 3844 (1995).
• E. Salinas and L. F. Abbott, A model of multiplicative responses in parietal cortex, PNAS, 93, 11956-61, (1996).
• Lecture 14 (T 10/29) (realvideo)
• Rehash of midterm exam.
• Lecture 15 (Th 10/31) (realvideo)
• Vector quantization (VQ).
• Principal component analysis (PCA).
• Assignment 7: Nonlinear network theory again.
• No class because of SFN meeting.
• Lecture 16 (Th 11/7) (Postscript, PDF, realvideo)
• Models of associative memory.
• J.J. Hopfield, Neural networks and physical systems with emergent collective computational abilities, Proc. Natl. Acad. Sci. USA, 79, 2554-58 (1982).
• M. V. Tsodyks and M. Feigelman.
• Summary of Lyapunov analysis: Chapter 3 from Slotine,1991
• Lecture 17 (T 11/12) (Postscript, PDF, realvideo)
• Delay activity. Griniasty-Tsodyks-Amit model.
• Y. Miyashita, Neuronal correlate of visual associative long-term memory in the primate temporal cortex, Nature, 335, 817-20 (1988).
• M. Griniasty, M. V. Tsodyks, and D. J. Amit, Conversion of temporal correlations between stimuli to spatial correlations between attractors, Neural Comput., 5, 1-17 (1993).
• D. J. Amit, N. Brunel, and M. V. Tsodyks, Correlations of cortical Hebbian reverberations: theory versus experiment, J. Neurosci., 14, 6435-45 (1994).
• D. J. Amit, The Hebbian paradigm reintegrated: local reverberations as internal representations, Behav. Brain Sci., 18, 617-26 (1995).
• D. Zipser, B. Kehoe, G. Littlewort, J. Fuster, A Spiking Network Model of Short-Term Active Memory, J. Neurosci., 13, 3406-3420 (1993)
• Assignment 8: VQ
• Lecture 18 (Th 11/14) (realvideo)
• Neural integrators
• H. S. Seung, How the brain keeps the eyes still, Proc. Natl. Acad. Sci. USA, 93:13339-44 (1996).
• K. Zhang, Representation of spatial orientation by the intrinsic dynamics of the head-direction cell ensemble: a theory,  J. Neurosci., 16, 2112-26 (1996).
• M. Camperi and X.-J. Wang. A model of visuospatial working memory in prefrontal cortex: recurrent network and cellular bistability.  J. Comput. Neurosci. 5, 383-405 (1998).
• Lecture 19 (T 11/19) (realvideo)
• Contrastive Hebbian learning and recurrent backprop learning
• Hertz, Krogh, and Palmer
• J.E.Slotine, W. Li, Applied Nonlinear Control, Chapter 3, Sections 3.1, 3.2, 3.4, 40-47, 47-53, 57-76, Prentice Hall, Inc., Englewood Cliffs, New Jersey (1991)
• J.J. Hopfield, Neurons with graded response have collective computational properties like those of two-state neurons, Proc. Natl. Acad. Sci. USA, 81, 3088-3092 (1984).
• Assignment 9 due.
• Lecture 20 (Th 11/21) (realvideo)
• Reinforcement learning. Hedonistic synapses.
• Lecture 21 (T 11/26) (realvideo)
• REINFORCE algorithms. Hedonistic neurons.
• Williams 92
• Assignment 10 due.
• Lecture 22 (T 12/3) (Postscript, PDF, realvideo)
• Gradient learning of trajectories: backpropagation and real-time recurrent learning.
• Hertz, Krogh & Palmer, pp.
• B. A. Pearlmutter. Gradient calculation for dynamic recurrent neural networks: a survey. IEEE Transactions on Neural Networks, 6(5):1212-1228 (1995).
• R. J. Williams and D. Zipser, Experimental Analysis of the Real-time Recurrent Learning Algorithm, Connection Science, 1, 87-111 (1989).
• B. A. Pearlmutter, Learning State Space Trajectories in Recurrent Neural Networks, Neural Computation, 1, 263-9 (1989).
• Lecture 23 (Th 12/5) (realvideo)
• Assignment 11 due.
• Lecture 24 (T 12/12) (realvideo)
• final review
• Final Exam Review Session (Th 12/18) (realvideo: part 1, part 2)