Surround modulation (Series et al. 2003) Modeling Adult Visual - PowerPoint PPT Presentation

Surround modulation (Series et al. 2003) Modeling Adult Visual Function Dr. James A. Bednar jbednar@inf.ed.ac.uk http://homepages.inf.ed.ac.uk/jbednar Apparent contrast Detection facilitated or Contour pops out reduces inhibited Many types of contextual interactions are known CNV Spring 2015: Modeling adult function 1 CNV Spring 2015: Modeling adult function 2 Surround modulation Proposed model circuit From Schwabe Effects depend et al. (2006): strongly on High-threshold contrast (Hirsch & Gilbert inhibitory 1991), (Weliky et al. 1995) interneurons (Schwabe et al. 2006) and on distance (Schwabe et al. 2006) Long-range (Angelucci & Bressloff 2006) excitatory lateral Distance-related connections effects match Long-range both lateral and excitatory feedback feedback connections connections CNV Spring 2015: Modeling adult function 3 CNV Spring 2015: Modeling adult function 4

GCAL-based SM model SM model size tuning • GCAL-based circuit for surround modulation V1 L2/3I • Separate inhibitory interneurons (Antolik 2010) • Long-range excitatory V1 L2/3E lateral connections • Separate simple and complex cell layers V1 L4 • No feedback connections; not published yet Single-unit response to larger patterns typically increases, (Antolik 2010; Antolik & Bednar 2015) (Philipp Rudiger) then decreases as inhibition is recruited CNV Spring 2015: Modeling adult function 5 CNV Spring 2015: Modeling adult function 6 Diversity in size tuning Diversity in OCTC tuning (Antolik 2010) (Antolik 2010) Model matches both typical and unusual orientation-contrast tuning types Model matches both typical and unusual size tuning responses CNV Spring 2015: Modeling adult function 7 CNV Spring 2015: Modeling adult function 8

The Tilt Aftereffect (TAE) Measuring perceived orientation Average: o OR preference = −3.0 Activation Neuron 2: OR preference = 0.0 o Activation = 1.0 Neuron 3: o OR preference = −16.8 Neuron 1: Activation = 0.6 = +22.5 o OR preference Activation = 0.24 • Assumption: perception based on population average • Bias in orientation perception after prolonged exposure • Vector average good for cyclic quantities • Allows model structure to be related to adult function • Use average to decode perception, before and after • Classic explanation: “fatigue” – activated neurons get tired, shifting the population average away adaptation CNV Spring 2015: Modeling adult function 9 CNV Spring 2015: Modeling adult function 10 TAE in Humans and LISSOM TAE Adaptation in LISSOM o • Null at zero : More 4 Aftereffect Magnitude inhibition, but no net • Direct effect for Adaptation + o 2 change in perception small angles − 0 ◦ • Direct effect: More o 0 • Indirect effect for inhibition for angles < 10 ◦ larger angles o −2 + Direct – Perception shifts − • Null effect at from 10 to 14 ◦ o −4 10 ◦ training angle o o o o o o o −90 −60 −30 0 30 60 90 • Indirect effect: Less Angle on Retina inhibition for angles < 60 ◦ Indirect + • Human, model − – Perception shifts match closely Mitchell & Muir 1976 60 ◦ from 60 to 58 ◦ HLISSOM • Due to synapses, not Input Histogram V1 Activity pattern difference tired neurons! CNV Spring 2015: Modeling adult function 11 CNV Spring 2015: Modeling adult function 12

McCollough effect test pattern Adaptation pattern Before adaptation, this pattern should appear monochrome Stare alternately at the two patterns for 3 minutes, moving your gaze to avoid developing strong afterimages CNV Spring 2015: Modeling adult function 13 CNV Spring 2015: Modeling adult function 14 McCollough effect McCollough effect: data • Effect measured in (McCollough 1965) humans at each After adaptation: angle between (Ellis 1977) • Vertical bars adaptation and test should be slightly • Strength falls off magenta smoothly with angle (Landisman & Ts’o 2002) • Horizontal bars should be slightly • V1 is earliest green possible substrate – first area showing • The effect should reverse if you tilt your head 90 ◦ , OR selectivity; has and disappear if you tilt 45 ◦ . 2.3 × 5.3mm macaque V1 color map CNV Spring 2015: Modeling adult function 15 CNV Spring 2015: Modeling adult function 16

LISSOM RG Color V1 Model LISSOM OR + Color map • Input: RGB images V1 • Decomposed into (Bednar et al. 2005) LGN Red, Green or Red, Green, Blue Luminosity channels Green/Red (e.g. no blue in (Bednar et al. 2005) Red/Green central fovea, ON OFF Calkins 2001 ) Retina • Processed by Red Green • Orientation map similar to animal maps Channel Channel color opponent • Color-selective cells occur in blobs Color Image retinal ganglia • Needs study of preferences of neurons in each blob CNV Spring 2015: Modeling adult function 17 CNV Spring 2015: Modeling adult function 18 Calculating McCollough Effect Model McCollough Effect • Perceived color estimated as a vector average of all units 6 4 • Vector direction: + for red-selective units, - for strength of the ME (in the model) 2 green-selective units 0 • Weighted by activation level and amount of color −2 selectivity −4 Result is a number from extreme red (positive) to extreme −6 green (negative), with approximately 0 being −45 −30 −15 0 15 30 45 60 75 90 105 120 135 orientation of the test pattern monochrome. CNV Spring 2015: Modeling adult function 19 CNV Spring 2015: Modeling adult function 20

Compared with humans Summary • GCAL can be compatible with actual circuit 1.2 simulated ME human data • Reproduces surprising features of surround modulation 1 • Afterffects arise from Hebbian adaptation of lateral 0.8 strength of the ME connections 0.6 0.4 • The same self-organizing processes can drive both 0.2 development and adaptation: both structure and function 0 • Novel prediction: Indirect effect due to weight −0.2 −45 −30 −15 0 15 30 45 normalization orientation of the test pattern • Project: details of wiring for inverted Mexican Hat CNV Spring 2015: Modeling adult function 21 CNV Spring 2015: Modeling adult function 22 McCollough Effect References Angelucci, A., & Bressloff, P . C. (2006). Contribution of feedforward, lateral and feedback connections to the classical receptive field center and extra- classical receptive field surround of primate V1 neurons. Progress in Brain Research , 154 , 93–120. Is the effect still Antolik, J. (2010). Unified Developmental Model of Maps, Complex Cells and present? Surround Modulation in the Primary Visual Cortex . Doctoral Dissertation, School of Informatics, The University of Edinburgh, UK. Antolik, J., & Bednar, J. A. (2015). A unified developmental model of maps, com- plex cells and surround modulation in the primary visual cortex. In prepa- ration. CNV Spring 2015: Modeling adult function 23 CNV Spring 2015: Modeling adult function 23

Bednar, J. A., De Paula, J. B., & Miikkulainen, R. (2005). Self-organization of Relationships to ocular dominance, cytochrome oxidase, and orientation. color opponent receptive fields and laterally connected orientation maps. Journal of Neurophysiology , 87 (6), 3126–3137. Neurocomputing , 65–66 , 69–76. McCollough, C. (1965). Color adaptation of edge-detectors in the human visual Calkins, D. J. (2001). Seeing with S cones. Progress in Retinal and Eye Research , system. Science , 149 (3688), 1115–1116. 20 (3), 255–287. Mitchell, D. E., & Muir, D. W. (1976). Does the tilt aftereffect occur in the oblique Ellis, S. R. (1977). Orientation selectivity of the McCollough effect: Analysis by meridian?. Vision Research , 16 , 609–613. equivalent contrast transformation. Perception and Psychophysics , 22 (6), Schwabe, L., Obermayer, K., Angelucci, A., & Bressloff, P . C. (2006). The role of 539–544. feedback in shaping the extra-classical receptive field of cortical neurons: Hirsch, J. A., & Gilbert, C. D. (1991). Synaptic physiology of horizontal connections A recurrent network model. The Journal of Neuroscience , 26 (36), 9117– in the cat’s visual cortex. The Journal of Neuroscience , 11 , 1800–1809. 9129. Landisman, C. E., & Ts’o, D. Y. (2002). Color processing in macaque striate cortex: Series, P ., Lorenceau, J., & Fregnac, Y. (2003). The “silent” surround of V1 recep- CNV Spring 2015: Modeling adult function 23 CNV Spring 2015: Modeling adult function 23 tive fields: Theory and experiments. Journal of Physiology (Paris) , 97 (4– 6), 453–474. Weliky, M., Kandler, K., Fitzpatrick, D., & Katz, L. C. (1995). Patterns of excitation and inhibition evoked by horizontal connections in visual cortex share a common relationship to orientation columns. Neuron , 15 , 541–552. CNV Spring 2015: Modeling adult function 23