Preparation for Informal Chat Flashcards
Paper I read to him was called
Pattern of neural responses in face regions are predicted by low-level image properities (2018)
In 2018 Paper
neural models of face processing by Haxby et al showed that diiferent regions are involved in processing different info from face - what regions? (3)
- Occipital face area (OFA)
- Fusiform face area (FFA)
- Superior temporal sulcus (STS)
In 2018 Paper
connection from OFA and STS is to due with
dynamic changes in face (e.g., expression)
In 2018 paper
connection of OFA and FFA is to do with
invariant (not-changing) features of face (e.g., identity)
In 2018 paper
evidence supporting OFA, FFA, STS is important for high-level attributes (e.g., identity and expression)
but this paper wanted to investigate whether
low-level properities (e.g., viewpoint) can explain the topographical organisation of these face regions
In 2018, experiment 1 involved
3 different viewpoints of familiar celeb faces
In 2018, experiment 2 involved
3 different viewpoints of different unfamiliar identities with 3 expressions = happy, disgust, fear
What was the conclusion of the 2018 paper?
Face region not exclusively dealing with high-level attributes but low level as well (image representation)
in fMRI, more blood flow to the part of the brain that is
most active
in fMRI we can measure the BOLD signal which stands for … to measure - (2)
Blood 02 Level Dependent
changes in oxygenation
Whole region of the brain is decided up into vowels (units of 3D image of brian) which represent
groups of neurons and its gross energy
What happens in univarate analysis for an fMRI experiment if we expose participants to face stimuli?
- We measuring the BOLD signla at each vowel in fMRI experiment
- Design metrics of the experiment is x (e.g., time duration of [facial] stimuli) and y is the BOLD signal of a voxel
- Put through regression model and get beta value of each voxel (how much is relation between x and y [ e.g, voxel ha shigh beta value when seeing face)
What happens in MVPA? - (3)
- Not looking at one voxel at the time
- Looking at a pattern of activation in cluster of voxels
- You train a classifer and then validate (e.g,, check if it can classify identity given a pattern of voxel)
What does MVPA stand for?
Multivoxel Pattern Analysis
Due to the curvature of the retina, left visual field…
stimulate right eye etc..