Blohm Lab Wiki

Latest revision as of 17:16, 13 August 2018

CoSMo logo

This page contains course materials for the CoSMo 2018 summer school.

Introduction

Jul 30 - Aug 4
Lecturers: Gunnar Blohm, Paul Schrater, Konrad Kording

Day 1 - Overview of modeling in neuroscience

CoSMo 2018 organizational slides

Konrad's and Gunnar's model pitches
Paul's multiple learning pitch
Paul's optimal forgetting pitch
Paul's deep learning bottleneck pitch
Paul's minimum intervention principle pitch

Afternoon tutorial 1: plotting neural data

Here is the file [Stevenson Data Set] As part of the tuning curve exercise we will understand it.

Tutorial is available here

Afternoon tutorial 2: gain modulation for reference frame transformations

The goal of this tutorial is to understand how gain modulation can be used for reference frame transformations and how gain modulation can emerge from training a simple artificial neural network carrying out reference frame transformations.
There are 2 different approaches to solving this:

exact determination of read-out weights from eye-position gain-modulated neurons as in this seminal paper. Here the solution can be found by computing the least-square optimal set of weights mapping the gain-modulated neurons (population code) to head-centered output neuron(s). For this to work, population code neurons need to be of the exponential function family.
training a neural network to perform reference frame transformations using this code. For this you can plot each individual neuron's receptive field for different eye positions and analyze how the receptive field changes with eye position in each network layer.

Day 2 - Bayesian approaches

Bayesian perception - an introduction: a tremendous book written by Wei Ji Ma, Konrad Kording, Daniel Goldreich

Morning lectures & tutorial

Konrad's Bayesian decoding and multi-sensory integration tutorials are available here (in folder day 2)

Afternoon lectures & tutorials
Dropbox link to Paul's slides and tutorial (in sub-folder decision_tutorial)

How to model tutorial
Modeling 101 slides

Day 3 - Linear systems and Kalman filtering

Morning: Linear systems (saccades)
Linear systems theory lecture
Eye movement tutorial and a possible solution for both time and frequency domain modelling
van Opstal syllabus - linear systems theory: a great syllabus developed by John van Opstal for CoSMo on using linear systems to model gaze control with theory, exercises and answers to exercises

Afternoon: Kalman filtering
Dropbox link to Paul's slides and tutorial (in sub-folder Kalman_lecture_tutorial)

Day 4 - Motor control

Morning: optimal control
Control slides
Matlab control tutorial files

Afternoon: Paper writing 101
Individual abstract writing document
Konrad's PLoS CB 10 simple rules paper on how to structure papers

Evening: causality in neuroscience
Konrad's discussion slides and tutorials on causality in neuroscience can be found here (in sub-folder causality)

Day 5 - Optimality

Can a neuroscientist understand a mirco-chip?
Konrad's article discussing this is published here

BADS model fitting
Lecturer: Luigi Acerbi

Luigi's slides
Here is the link to Luigi's tutorial files on Github.

Day 6 - machine learning

Konrad's tutorial on machine learning for neuroscience can be found here (in sub-folder ML)

Motor control

Aug 6 - 7
Lecturers: Alaa Ahmed, Frederic Crevecoeur, Reza Shadmehr

Alaa's and Reza's first day materials
Reza's neural prelude to movement slides
Alaa's PM tutorial
Reza's neural coding in the cerebellum slides
Reza's reaction time slides
Alaa's slides (all in one deck)

Fred's control tutorial
Lecture slides and further readings
Tutorial instructions and associated Matlab script

The Bayesian Brain

Aug 8 - 9
Lecturers: Megan Peters & Larry Maloney

Day 1
Larry's introductory lecture
Larry's first tutorial
Second lecture from Larry
Second set of exercises from Larry

Day 2
Larry's third lecture slides
Larry's 4th slide deck
Larry's last set of slides

All of Megan's slides and tutorial materials for both days can be found here!

Open Science & other discussions

Aug 9

Open science and strategies to become a modeller slides. YOU CAN DO IT!!!

Group Projects

Jul 30 - Aug 11

Instructions: Every group will have a 30min slot (20min presentation, 10min questions). The research question, hypotheses and rationale for the choice of the approach should be clearly presented. Models, simulations, results, discussion etc should be detailed enough for everyone to follow.

Best group gets a free 10min talk at Advances in Motor Control and Motor Learning 2018 (SfN satellite workshop)!!! (confirmed by John Krakauer & Maurice Smith)
The winner has to apply too and specify you are CoSMo 2018 project winner...

PROJECT PRESENTATIONS (Sat, Aug 11)
Presentations will take place 1-6pm in Bruininks Hall Room 512B
Presentation template PowerPoint file

1 - 1:30pm | Fisher Information-Information Fishers (FIIFs): Eva Berlot | Noah Steinberg | Tyler Manning | Lina Koronfel
Project title: "Analyzing the Effect of Noise Correlations on Decoding Neuronal Recording"
Abstract: A common idea in neuroscience is that neuronal populations faithfully encode the relevant stimuli. While attention has been paid to the influence of tuning properties on encoding performance, we do not know how the structure of noise correlations relates to information encoding. To investigate this, we have simulated a population of leaky integrate-and-fire neurons, and examined how modulating noise and stimulus correlations affects decoding accuracy. We find decoding is highest for a population with the lowest noise correlations. It is thus important to consider tuning properties in addition to noise correlation jointly when examining how neurons encode stimuli.
Presentation slides

1:30 - 2pm | Elucidating Attribution Given Errors By Estimating A Vector Regression Solution (EAGER BEAVeRS) : Anjana Gayathri Arunachalam | Wendy Boehm | Jen Ruttle | Chris Yang
Project title: "Modeling Error Source Attribution During Motor Adaptation"
Abstract: Humans can maintain accurate motor performance in the face of motor errors, regardless of whether the errors originate internally from their own body or externally from the world. Currently, little is known about how the brain identifies the errors’ source and adapts control accordingly. We designed a linear regressor to discriminate the source of errors in a simulated force field adaptation task. The model is capable of updating body and world-centric parameters driven by internally- and externally-attributed errors, respectively. This model suggests a possible mechanism for attributing a motor error to an internal or external source.
Presentation slides

2 - 2:30pm | Keep Kalm-an Carry On: Arthur-Ervin Avramiea | Francisco Sacadura | Sohrab Salimian | Nicola Popp | Gwydion Williams
Project title: "What determines the relationship between motor variability and the speed of motor learning?"
Abstract: Motor variability has been associated with enhanced motor learning, though recent findings question the generality of this effect and the task features that allow this facilitation require clarification. We approximated the optimal learner during a visuomotor learning task (using a Kalman Filter), introducing a perturbation which required learning to maintain optimal performance. We varied features of the model and of the task, measuring the relationship between motor variability and learning for each iteration. Our results show that the relationship between learning and motor variability is determined by the features of the task in question, rather than being a general phenomenon.
Presentation slides

2:30 - 3pm | How I match your model (HIMYM): Lihi Gibor | Giovanni Martino | Justinas Česonis | Michael Joch
Project title: "Modeling feedback gains in a perturbed reaching task"
Abstract: In perturbed reaching tasks, the visuomotor reflex reaction strength has been shown to decrease near the end of movement. However, it remains unclear whether this decrease can be explained by system delays or distance-to-target. Here we use an optimal feedback controller to model reaching movements towards a target, perturbed at different positions. Our results show that system delays, rather than distance-to-target, modulate the visuomotor reflex reaction strength. This suggests that the reflexive reaction strength may be dependent on movement kinematics rather than being pre-set for a given movement.
Presentation slides

3 - 3:30pm | Pupil Orientation Meets Depth Perception (POMDP) Team: Edoardo Balzani | Nathanael Larigaldie | Gabor Lengyel | Jean-Paul Noel | Majed Samad
Project title: "Hysteresis in stereoscopic fusion: A phenomenon with a surprising amount of depth"
Abstract: If each of our eyes sees a line and these are sufficiently close, we can fuse them into a 3D-percept. Research has shown that this phenomenon is history dependent: if we fused signals in the recent past we are more likely to fuse them now. However, whether this hysteresis effect is stimulus-dependent is unknown. Here we show that both a sequential causal inference model and a switching Kalman filter framework predict that hysteresis for fusion/fission scales with a prior for continuity. This prior is built sequentially given stimulus statistics. This suggests that stereo vision is attuned to the statistics of the environment.
WINNER !!!!!
Presentation slides and GUI for simulations

3:30 - 4pm | Smooth Optimator5 (S.O.5): Clara Cámara | Benjamin Cuthbert | Stephan Dobri | Tom Nissens | Marta Russo
Project title: "Not So Smooth: Modelling Reaching Behaviour of Children with Developmental Coordination Disorder"
Abstract: Reaching movements in healthy subjects usually display one-peak speed profiles. Children with Developmental Coordination Disorder (DCD) instead exhibit intermittent trajectories with multiple peaks in the speed profile. The underlying mechanism is still debated. Here we investigated whether a violation in the internal model of the arm dynamics can cause the observed characteristics of DCD, by means of Optimal Feedback Control (OFC). We found that small changes in the internal model resulted in curved paths and rough speed profiles. We suggest a new hypothesis for DCD deficits: multiple speed peaks could be explained by an inaccurate internal model of arm dynamics.
Presentation slides

4 - 4:30pm | SWoNs: Working on Networks (SWoNs): Habiba Azad, Brandon Caie, Matt Laporte, Richard Moulton, and Akis Stavropoulos
Project title: "Synchronizing to Decide in Small-World Networks"
Abstract: Network neuroscience assumes that synchronous activity across the brain reflects coordination between distinct regions. Although a network’s connectivity structure determines its ability to maintain temporal patterns in the presence of noise, it is unclear how this specifically leads to decisions. To explore this, we model value-based decisions where two oscillating networks suppress each other’s capacity to synchronize. We show the emergence of winner-take-all dynamics and demonstrate that small-world structure provides an advantage in the competition for synchrony in the presence of noise. We conclude that decisions could result from connectivity-dependent competition for synchronous activity between regions of the brain.
Presentation slides

4:30 - 5pm | Confessions of a Sliding Cube (C.O.A.S.C): Andreea Loredanna Cretu | Harun Karimpur | Padmapriya Muralidharan | Sabrina Hansmann-Roth
Project title: "To slide or not to slide: Deviations of friction estimation in intuitive physics"
Abstract: At a glance, we perceive whether the ground is icy or safe to walk on with an intuitive understanding about the physical properties of objects. Here, we investigate how people weight object size and friction in their interpretations. Participants passively view a video of a cube on a rotating plane. They are then asked to tilt the plane to the maximum estimated angle at which they expect the object to start sliding. Our results highlight a common misconception of intuitive physics demonstrating a deviance from an ideal observer model which predicts that size does not play a role for friction estimation.
Presentation slides

5 - 5:30pm | Better Call Paul (B.C.P.): Jana Masselink | Serena Ricci | Judith Rudolph | Marion Forano
Project title: "Does motor variability influence learning in uncertain situations?"
Abstract: We seem to benefit from variability introduced by our own motor system, but experiencing environmental uncertainty reduces our ability to learn. However, it has not yet been investigated how motor variability modulates the effect of environmental uncertainty on visuomotor adaptation. Here we use an adaptive Kalman filter to predict adaptation behavior under different perturbation uncertainties. We show that for higher motor noise and perturbation uncertainty, adaptation rate and final compensation increase, and difference in motor variance decreases. Therefore, this study shows that larger variabilities in movements aid learning under greater uncertainty.
Presentation slides

5:30 - 6pm | Face Adaptation Networks (FANs): Sunwoo Kwon | Tina Liu | Cindy Tu | Nils Yang
Project title: "Perception of face gender under ambiguity: neural network of face adaptation"
Abstract: Our perception of the gender of a face is dependent on the recent gender history. Face adaptation studies suggest that constant exposure to one gender shifts the perception to the opposite gender. However, we are currently lacking models that can successfully describe this behavioral phenomenon. We thus implement a neural network model of face gender perception. We find that this model faithfully captures the dependency of face perception on gender history. This demonstrates that perception of gender is influenced by the frequency of exposure to gender faces.
Presentation slides

Project voting form (requires Google sign-in to prevent duplicates - no user info recorded, it's an anonymous vote)

6pm +: Closing reception @ Loring Pasta bar

@@ Line 83: / Line 83: @@
 ''' BADS model fitting ''' <br>
 Lecturer: Luigi Acerbi <br>
+[[Media:Acerbi-CoSMo2018.pdf | Luigi's slides]] <br>
+[https://github.com/lacerbi/cosmo-2018-tutorial Here is the link] to Luigi's tutorial files on Github. <br>
+=== Day 6 - machine learning ===
+Konrad's tutorial on machine learning for neuroscience can be found [https://www.dropbox.com/sh/mqr7x1q8rk9129h/AAChwxoOwvQ3Y6U47cO8HRkVa?dl=0 here] (in sub-folder ML) <br>
+----
+== Motor control ==
+''Aug 6 - 7'' <br>
+Lecturers: Alaa Ahmed, Frederic Crevecoeur, Reza Shadmehr <br>
+''' Alaa's and Reza's first day materials ''' <br>
+[[Media:Neural_prelude_to_movement_Reza_CoSMo2018.pdf | Reza's neural prelude to movement slides]] <br>
+[[Media:Assignment_Alaa_CoSMo2018.zip | Alaa's PM tutorial]] <br>
+[[Media:Neural_coding_in_the_cerebellum_Reza_CoSMo2018.pdf | Reza's neural coding in the cerebellum slides]] <br>
+[[Media:Reaction_time_Reza_CoSMo2018.pdf | Reza's reaction time slides]] <br>
+[[Media:Ahmedfinal_slides2_CoSMo2018.pdf | Alaa's slides]] (all in one deck) <br>
+''' Fred's control tutorial ''' <br>
+[https://drive.google.com/file/d/1KTvCCXhPGvh8NJSy4Fdsajh1hxc132qR/view Lecture slides] and [[Media:CoSMo2018_controllit.zip | further readings]] <br>
+[[Media:Assignment_CoSMo2018.pdf | Tutorial instructions]] and [[Media:Control_script_CoSMo2018.m | associated Matlab script]] <br>
+----
+== The Bayesian Brain ==
+''Aug 8 - 9'' <br>
+Lecturers: Megan Peters & Larry Maloney <br>
+''' Day 1''' <br>
+[[Media:CoSMo2018_Module01_Lecture_Larry.pdf | Larry's introductory lecture]] <br>
+[[Media:CoSMo_Exercises_Weds01_Larry_CoSMo2018.pdf | Larry's first tutorial]] <br>
+[[Media:CoSMo2018_Module02_Lecture_Larry.pdf | Second lecture from Larry]] <br>
+[[Media:CoSMo2018_Exercises_Larry02.pdf | Second set of exercises from Larry]] <br>
+''' Day 2 ''' <br>
+[[Media:CoSMo2018_Module03_Lecture_Larry.pdf | Larry's third lecture slides]] <br>
+[[Media:CoSMo2018_Module04_Lecture_Larry.pdf | Larry's 4th slide deck]] <br>
+[[Media:CoSMo2018_Module05_Lecture_Larry.pdf | Larry's last set of slides]] <br>
+[https://drive.google.com/drive/folders/1XFtyhB8OeAGQbPI4o7wBkqjXSg28EMlz All of Megan's slides and tutorial materials for both days can be found here!] <br>
+----
+== Open Science & other discussions ==
+''Aug 9'' <br>
+[[Media:OpenScience_CoSMO2018.pdf | Open science and strategies to become a modeller slides]]. YOU CAN DO IT!!! <br>
+----
+== Group Projects ==
+''Jul 30 - Aug 11'' <br>
+'''Instructions:''' Every group will have a 30min slot (20min presentation, 10min questions). The research question, hypotheses and rationale for the choice of the approach should be clearly presented. Models, simulations, results, discussion etc should be detailed enough for everyone to follow. <br>
+'''Best group gets a free 10min talk at Advances in Motor Control and Motor Learning 2018 (SfN satellite workshop)!!!''' (confirmed by John Krakauer & Maurice Smith) <br>
+The winner has to apply too and specify you are CoSMo 2018 project winner... <br>
+'''PROJECT PRESENTATIONS (Sat, Aug 11)''' <br>
+''Presentations will take place 1-6pm in Bruininks Hall Room 512B'' <br>
+[[Media:GroupProjectPresentationTemplate2018.pptx | Presentation template]] PowerPoint file <br>
+''1 - 1:30pm'' |
+'''Fisher Information-Information Fishers (FIIFs)''': Eva Berlot | Noah Steinberg | Tyler Manning | Lina Koronfel <br>
+'''Project title''': "Analyzing the Effect of Noise Correlations on Decoding Neuronal Recording" <br>
+'''Abstract''': A common idea in neuroscience is that neuronal populations faithfully encode the relevant stimuli. While attention has been paid to the influence of tuning properties on encoding performance, we do not know how the structure of noise correlations relates to information encoding. To investigate this, we have simulated a population of leaky integrate-and-fire neurons, and examined how modulating noise and stimulus correlations affects decoding accuracy. We find decoding is highest for a population with the lowest noise correlations. It is thus important to consider tuning properties in addition to noise correlation jointly when examining how neurons encode stimuli. <br>
+[[Media:CoSMO_GroupProjectPresentation_FIIF_2018.pdf | Presentation slides]] <br>
+''1:30 - 2pm'' |
+'''Elucidating Attribution Given Errors By Estimating A Vector Regression Solution (EAGER BEAVeRS)''' : Anjana Gayathri Arunachalam | Wendy Boehm | Jen Ruttle | Chris Yang<br>
+'''Project title''': "Modeling Error Source Attribution During Motor Adaptation" <br>
+'''Abstract''': Humans can maintain accurate motor performance in the face of motor errors, regardless of whether the errors originate internally from their own body or externally from the world. Currently, little is known about how the brain identifies the errors’ source and adapts control accordingly. We designed a linear regressor to discriminate the source of errors in a simulated force field adaptation task. The model is capable of updating body and world-centric parameters driven by internally- and externally-attributed errors, respectively. This model suggests a possible mechanism for attributing a motor error to an internal or external source. <br>
+[[Media:CoSMo2018_EAGEr_BEAVeRS.pdf | Presentation slides]] <br>
+''2 - 2:30pm'' |
+'''Keep Kalm-an Carry On''': Arthur-Ervin Avramiea | Francisco Sacadura | Sohrab Salimian | Nicola Popp | Gwydion Williams <br>
+'''Project title''': "What determines the relationship between motor variability and the speed of motor learning?" <br>
+'''Abstract''': Motor variability has been associated with enhanced motor learning, though recent findings question the generality of this effect and the task features that allow this facilitation require clarification. We approximated the optimal learner during a visuomotor learning task (using a Kalman Filter), introducing a perturbation which required learning to maintain optimal performance. We varied features of the model and of the task, measuring the relationship between motor variability and learning for each iteration. Our results show that the relationship between learning and motor variability is determined by the features of the task in question, rather than being a general phenomenon. <br>
+[[Media:FinalPresentation_KeepKalm-an_carry_on.pdf | Presentation slides]] <br>
+''2:30 - 3pm'' |
+'''How I match your model (HIMYM)''': Lihi Gibor | Giovanni Martino | Justinas Česonis | Michael Joch <br>
+'''Project title''': "Modeling feedback gains in a perturbed reaching task" <br>
+'''Abstract''': In perturbed reaching tasks, the visuomotor reflex reaction strength has been shown to decrease near the end of movement. However, it remains unclear whether this decrease can be explained by system delays or distance-to-target. Here we use an optimal feedback controller to model reaching movements towards a target, perturbed at different positions. Our results show that system delays, rather than distance-to-target, modulate the visuomotor reflex reaction strength. This suggests that the reflexive reaction strength may be dependent on movement kinematics rather than being pre-set for a given movement. <br>
+[[:Media:Modeling_feedback_gains_in_a_perturbed_reaching_task.pdf | Presentation slides]] <br>
+'' 3 - 3:30pm'' |
+'''Pupil Orientation Meets Depth Perception (POMDP) Team''': Edoardo Balzani | Nathanael Larigaldie | Gabor Lengyel | Jean-Paul Noel | Majed Samad <br>
+'''Project title''': "Hysteresis in stereoscopic fusion: A phenomenon with a surprising amount of depth" <br>
+'''Abstract''': If each of our eyes sees a line and these are sufficiently close, we can fuse them into a 3D-percept. Research has shown that this phenomenon is history dependent: if we fused signals in the recent past we are more likely to fuse them now. However, whether this hysteresis effect is stimulus-dependent is unknown. Here we show that both a sequential causal inference model and a switching Kalman filter framework predict that hysteresis for fusion/fission scales with a prior for continuity. This prior is built sequentially given stimulus statistics. This suggests that stereo vision is attuned to the statistics of the environment. <br>
+'''WINNER !!!!!''' <br>
+[[Media:Presentation_correct_template_final_POMDP.pdf | Presentation slides]] and [[Media:Gui_POMDP_CoSMo2018.zip | GUI for simulations]] <br>
+''3:30 - 4pm'' |
+'''Smooth Optimator5 (S.O.5)''': Clara Cámara | Benjamin Cuthbert | Stephan Dobri | Tom Nissens | Marta Russo <br>
+'''Project title''': "Not So Smooth: Modelling Reaching Behaviour of Children with Developmental Coordination Disorder" <br>
+'''Abstract''': Reaching movements in healthy subjects usually display one-peak speed profiles. Children with Developmental Coordination Disorder (DCD) instead exhibit intermittent trajectories with multiple peaks in the speed profile. The underlying mechanism is still debated. Here we investigated whether a violation in the internal model of the arm dynamics can cause the observed characteristics of DCD, by means of Optimal Feedback Control (OFC). We found that small changes in the internal model resulted in curved paths and rough speed profiles. We suggest a new hypothesis for DCD deficits: multiple speed peaks could be explained by an inaccurate internal model of arm dynamics. <br>
+[[Media:Final_Presentation_no_videos_-_smooth_optimators5.pdf | Presentation slides]] <br>
+''4 - 4:30pm'' |
+'''SWoNs: Working on Networks (SWoNs)''':
+Habiba Azad, Brandon Caie, Matt Laporte, Richard Moulton, and Akis Stavropoulos <br>
+'''Project title''': "Synchronizing to Decide in Small-World Networks" <br>
+'''Abstract''': Network neuroscience assumes that synchronous activity across the brain reflects coordination between distinct regions. Although a network’s connectivity structure determines its ability to maintain temporal patterns in the presence of noise, it is unclear how this specifically leads to decisions. To explore this, we model value-based decisions where two oscillating networks suppress each other’s capacity to synchronize. We show the emergence of winner-take-all dynamics and demonstrate that small-world structure provides an advantage in the competition for synchrony in the presence of noise. We conclude that decisions could result from connectivity-dependent competition for synchronous activity between regions of the brain. <br>
+[[Media:SWoNs_presentation_Combined_CoSMo2018.pdf | Presentation slides]] <br>
+''4:30 - 5pm'' |
+'''Confessions of a Sliding Cube (C.O.A.S.C)''':
+Andreea Loredanna Cretu | Harun Karimpur | Padmapriya Muralidharan | Sabrina Hansmann-Roth <br>
+'''Project title''': "To slide or not to slide: Deviations of friction estimation in intuitive physics" <br>
+'''Abstract''': At a glance, we perceive whether the ground is icy or safe to walk on with an intuitive understanding about the physical properties of objects. Here, we investigate how people weight object size and friction in their interpretations. Participants passively view a video of a cube on a rotating plane. They are then asked to tilt the plane to the maximum estimated angle at which they expect the object to start sliding. Our results highlight a common misconception of intuitive physics demonstrating a deviance from an ideal observer model which predicts that size does not play a role for friction estimation. <br>
+[[Media:CoaSC_CoSMo2018.pdf | Presentation slides]] <br>
+''5 - 5:30pm'' |
+'''Better Call Paul (B.C.P.)''': Jana Masselink | Serena Ricci | Judith Rudolph | Marion Forano <br>
+'''Project title''': "Does motor variability influence learning in uncertain situations?" <br>
+'''Abstract''': We seem to benefit from variability introduced by our own motor system, but experiencing environmental uncertainty reduces our ability to learn. However, it has not yet been investigated how motor variability modulates the effect of environmental uncertainty on visuomotor adaptation. Here we use an adaptive Kalman filter to predict adaptation behavior under different perturbation uncertainties. We show that for higher motor noise and perturbation uncertainty, adaptation rate and final compensation increase, and difference in motor variance decreases. Therefore, this study shows that larger variabilities in movements aid learning under greater uncertainty. <br>
+[[Media:Cosmo2018_BCP.pdf | Presentation slides]] <br>
+''5:30 - 6pm'' |
+'''Face Adaptation Networks (FANs)''': Sunwoo Kwon | Tina Liu | Cindy Tu | Nils Yang <br>
+'''Project title''': "Perception of face gender under ambiguity: neural network of face adaptation" <br>
+'''Abstract''': Our perception of the gender of a face is dependent on the recent gender history. Face adaptation studies suggest that constant exposure to one gender shifts the perception to the opposite gender. However, we are currently lacking models that can successfully describe this behavioral phenomenon. We thus implement a neural network model of face gender perception. We find that this model faithfully captures the dependency of face perception on gender history. This demonstrates that perception of gender is influenced by the frequency of exposure to gender faces. <br>
+[https://docs.google.com/presentation/d/1jETKox8ah7o53oryDnqIEJMIV2cVMN543o5TN3M8D2c/edit?usp=sharing Presentation slides] <br>
+'''[https://goo.gl/forms/iBRGlpRlZmhYfAYx2 Project voting form]''' (requires Google sign-in to prevent duplicates - no user info recorded, it's an anonymous vote) <br>
+''6pm +'': '''Closing reception @ Loring Pasta bar''' <br>