This document contains topics indicated by the professor as important to know. If the professor made
no mention about things you should know, then you find: [no mention of important stuff]. This is not
to say that no important topics were discussed that lecture/module, just that the professor did not
make an explicit mention about it.
Lecture 1
V4 neurons like a specific combination of texture and shape.
0.5 x eccentricity = critical distance.
Module 1.1
Visual angle provides a measure of eccentricity (how far out into periphery the stimulus is),
regardless of distance to observer.
Visual angle to express size of object projection. This does change with distance.
Convergence affects receptive fields: spatial convergence is stronger for periphery
receptive fields are larger there.
3 reasons for seeing worse in the periphery:
1. More optical aberrations
2. Fewer receptors
3. Stronger spatial convergence, larger receptive fields
Module 1.2
[no mention of important stuff]
Lecture 2
Predictive coding theory: strongly predicted info is reduced.
Module 2.1
Feedback and feedforward creates recurrency: sustained activity through closed feedback
loops.
Classifying if its spatial context or temporal context, long-range or short range:
o Contextual: what type of context is influencing your perception.
Spatial context = current surroundings of object/stimulus you look at,
influences perception of object.
Temporal context = stuff you saw earlier influences current perception.
o Hierarchical: level in hierarchy it plays out.
Local, short-range: concrete stimuli, sensory, small receptive fields, detail,
short-time scale.
Global, long-range: abstract, semantic knowledge, basic gist of things rather
than detail, larger receptive fields, long time scales.
Module 2.2
Designing a CNN with feedforward (standard) or feedback + error signal, both sevelop filters
we see in human brain.
o Carry input feedforward
o Carry deviations from expectations
o Develop similar filters
Lecture 3
Things mentioned in: video, quizzes and lecture are crucial to get. Don’t have to understand
everything from the articles.
, Module 3.1
Polar coordinates: have two dimensions: 1) how far is stimulus out in terms of eccentricity,
2) polar angle: where around circle are we (which quadrant, or more precise)
Don’t have to know all regions of image below by heart, just know: various V regions in the
back, IPS to the front & that it has multiple maps, and also maps that go into temporal lobe.
Receptive field sizes increase as function of eccentricity and as level in hierarchy.
Don’t need to understand whole image below, just know: pRF mapping: you build a model
where you entre what the stimulus does (travelling bar from L R) = predictor. From that
visual bar, make a model of what activity should look like if it hits receptive field. Then get
data from the scanner and optimize the fit. Reconstruct from the timeline of activity which
part of space and which width the voxel have been responding.
Module 3.2
[no mention of important stuff]
Module 3.3
[no mention of important stuff]
Lecture 4 – NOT EXAM MATERIAL.
no mention about things you should know, then you find: [no mention of important stuff]. This is not
to say that no important topics were discussed that lecture/module, just that the professor did not
make an explicit mention about it.
Lecture 1
V4 neurons like a specific combination of texture and shape.
0.5 x eccentricity = critical distance.
Module 1.1
Visual angle provides a measure of eccentricity (how far out into periphery the stimulus is),
regardless of distance to observer.
Visual angle to express size of object projection. This does change with distance.
Convergence affects receptive fields: spatial convergence is stronger for periphery
receptive fields are larger there.
3 reasons for seeing worse in the periphery:
1. More optical aberrations
2. Fewer receptors
3. Stronger spatial convergence, larger receptive fields
Module 1.2
[no mention of important stuff]
Lecture 2
Predictive coding theory: strongly predicted info is reduced.
Module 2.1
Feedback and feedforward creates recurrency: sustained activity through closed feedback
loops.
Classifying if its spatial context or temporal context, long-range or short range:
o Contextual: what type of context is influencing your perception.
Spatial context = current surroundings of object/stimulus you look at,
influences perception of object.
Temporal context = stuff you saw earlier influences current perception.
o Hierarchical: level in hierarchy it plays out.
Local, short-range: concrete stimuli, sensory, small receptive fields, detail,
short-time scale.
Global, long-range: abstract, semantic knowledge, basic gist of things rather
than detail, larger receptive fields, long time scales.
Module 2.2
Designing a CNN with feedforward (standard) or feedback + error signal, both sevelop filters
we see in human brain.
o Carry input feedforward
o Carry deviations from expectations
o Develop similar filters
Lecture 3
Things mentioned in: video, quizzes and lecture are crucial to get. Don’t have to understand
everything from the articles.
, Module 3.1
Polar coordinates: have two dimensions: 1) how far is stimulus out in terms of eccentricity,
2) polar angle: where around circle are we (which quadrant, or more precise)
Don’t have to know all regions of image below by heart, just know: various V regions in the
back, IPS to the front & that it has multiple maps, and also maps that go into temporal lobe.
Receptive field sizes increase as function of eccentricity and as level in hierarchy.
Don’t need to understand whole image below, just know: pRF mapping: you build a model
where you entre what the stimulus does (travelling bar from L R) = predictor. From that
visual bar, make a model of what activity should look like if it hits receptive field. Then get
data from the scanner and optimize the fit. Reconstruct from the timeline of activity which
part of space and which width the voxel have been responding.
Module 3.2
[no mention of important stuff]
Module 3.3
[no mention of important stuff]
Lecture 4 – NOT EXAM MATERIAL.
