lecture 7 - encoding of real-world locations Flashcards

1
Q

receptive field

A
  • regions in a sensory input space (like the retina for vision or the skin for touch) where stimulation leads to activation of a neuron.
  • exists in sensory coordinates
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

sensory coordinates

A

how the brain represents sensory information

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

receptive fields in sensory coordinates

A
  1. RFs are localized on the surface of the sensory input organ
  2. RFs of sensory/abstract dimensions are well-described by gaussian models
  3. qualities of maps: RF Location-Size Correlations: RFs can vary in size and location depending on their sensory roles. Some RFs are broadly tuned, meaning they respond to a wide range of stimuli
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is the brain’s task regarding sensory input?

A
  • the brain has to infer the state of the world from sensory input
  1. all information enters the brain in sensory coordinates: e.g., sensory coordinates based on the position of light on the retina
  2. the brain needs to transform these into world coordinates
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

why does the brain need to transform sensory coordinates into world coordinates

A

To achieve a stable, global understanding of where objects are in the real world, not just their location on sensory organs like the retina.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What is an example of the brain using world coordinates?

A

Knowing a flower is in front of you, not just that its image fell on a specific part of your retina.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

the challenge of transforming sensory coordinates into world coordinates

A

sensory input is highly dynamic and subject to changes from movement.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

How does eye movement challenge sensory input processing?

A

Eye movements can completely jumble up incoming sensations, making it difficult to interpret stable visual information.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

How does the brain create world-centered representations?

A
  • By integrating across actions, time, modalities, and other factors.
  • how the brain performs this integration from low-levels to high-levels is not clear.
  • it is clear, however, that the brain does represent information in world-centred coordinates
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What did behaviorists believe about how rats learn to find food in a maze?

A

Behaviorists believed that rats learn to find food only by memorizing stimulus-response (reward) pairings.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What did Tolman discover about how rats learn mazes?

A

Tolman found that rats learn the layout of a maze even without rewards and would be faster at finding food in a maze they had already traversed earlier.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What does Tolman’s discovery suggest about rats’ learning?

A

It suggests that rats’ ability to find rewards in a maze is better explained by the idea of a cognitive map rather than simple stimulus-response learning.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is a cognitive map

A

A cognitive map is a mental map of the world around us, used to represent spatial relationships between objects in world-centered terms.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

How does the cognitive map differ from stimulus-response learning?

A

The cognitive map implies that animals (and humans) can represent spatial relationships in world-centered terms rather than relying solely on local cues.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

In what kind of coordinates do receptive fields exist?

A

sensory coordinates

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What is a place field?

A

A place field is a specific spatial location where neurons in the hippocampus fire when a rat is in that location, discovered by O’Keefe et al. in the 1970s.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What can a population of place cells do

A

Report the animal’s location in the surroundings to the rest of the brain

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

How does the hippocampus represent paths?

A
  • The hippocampus represents paths as a sequence of activations of place fields, allowing the reconstruction of space and time from hippocampal activity.
  • this means we can deduce where a mouse is from its hippocampus activation
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What happens during sleep in terms of hippocampal activity?

A

During sleep, the sequence of hippocampal activity is replayed in compressed time, either forward or backward (sped up or reversed).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

What is preplay, and what are its functions?

A

Preplay is hippocampal replay that occurs before events and serves two functions:

  1. Memory formation: Linking experiences for learning.
  2. Planning wakeful behaviors: we therefore can see what a rat is going to do based on the replay we see in the hippocampus
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What happens if the hippocampal replay mechanism is (task) inaccurate?

A

Taks-focused replay predicts accurate decisions, so if replay mechanism is inaccurate, decision-making will be inaccurate

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

How can disrupting replay affect behavior?

A

Disrupting hippocampal replay at decision points can reduce performance on tasks by impairing accurate decision-making.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

What do memory formation and wakeful behaviors together demonstrate?

A

They show that replay represents what the rat is thinking and is crucial for both learning and decision-making.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

What is remapping in place cells?

A

Remapping refers to how place cells adapt to track changes in surroundings by activating different place fields for different environments.

25
Q

Can a single neuron have multiple place fields?

A

Yes, a single neuron can have different place fields in different surroundings, allowing it to adapt without needing a unique place field for every location.

26
Q

What happens to the place field structure when a cell returns to a previous environment?

A

The cell picks up its previous place field structure when returning to the same environment.

27
Q

Why do we study hippocampal inputs and outputs?

A

To understand the mechanism for remapping by examining hippocampal place fields in relation to afferent (incoming) signals.

28
Q

Where do hippocampal inputs come from?

A

Hippocampal inputs come from the entorhinal cortex (EC).

29
Q

What is the role of the entorhinal cortex (EC)?

A

The EC acts as a gateway, transmitting spatial and sensory information from other brain regions to the hippocampus.

30
Q

Why is the entorhinal cortex important for place fields in the hippocampus?

A

Place fields in the hippocampus might emerge from the integration and transformation of inputs from the EC.

31
Q

What are grid cells in the entorhinal cortex (EC)?

A

Grid cells are neurons in the EC with world-coordinate receptive fields that fire in a grid-like pattern across multiple locations, helping animals track their position in the environment.

32
Q

How are grid cells different from hippocampal place cells?

A

Unlike hippocampal place cells, which fire at specific locations, grid cells fire at regularly spaced intervals across the entire environment.

33
Q

What pattern do grid cells form in space?

A

Grid cells form a hexagonal, crystalline structure in world-centered space.

34
Q

What properties can differ among EC grid cells?

A
  1. Frequency of repetition: size of the field
  2. Phase: location of the field
    - Grid cells with the same frequency can have offsets (different phases), meaning their activation peaks occur at different locations.
35
Q

What does the frequency of a grid cell define?

A

The frequency defines the size of the grid, indicating how far apart the hexagonal points of activation are.

  • Larger spacing between grid points (low frequency).
  • Smaller spacing between grid points (high frequency).
36
Q

What does the phase of a grid cell determine?

A

The phase determines the exact location of the grid points in space.

37
Q

What are the different bases for spatial coding?

A
  1. Gaussian basis: Used in sensory systems to encode locations in sensory coordinates (e.g., x and y).
  2. Fourier basis: Used in navigational systems to encode locations in world coordinates (e.g., phase and frequency).
38
Q

How do receptive fields in sensory systems encode spatial information?

A

They use a Gaussian basis to encode location in sensory coordinates (x and y).

39
Q

How do place fields in navigational systems encode spatial information?

A

They use a Fourier basis to encode location in world coordinates (phase and frequency).

40
Q

What is the significance of linear combinations in spatial coding?

A

Linear combinations of grid cells with specific weights can encode precise locations, either in sensory or world coordinates.

41
Q

What is the relationship between grid cells and place cells?

A

A single bank of grid cells can be reweighted to produce different place fields in the hippocampus.

42
Q

What did Doeller et al. aim to observe in their VR experiment on grid cells?

A

They aimed to observe if humans show grid-like periodic patterns of activation during spatial navigation.

43
Q

What is the hypothesis of the VR experiment on grid cells in humans?

A
  1. Strong activation when moving in your grid direction (over grid bumps).
  2. Weak activation when moving off-grid direction (over grid troughs).
44
Q

What prediction was made about grid cell firing in humans?

A

Grid cells were expected to show periodic peaks in activity every 60°, as they move in directions aligned with grid bumps.

45
Q

What were the key results of Doeller et al.’s VR experiment?

A
  1. Voxels with 60° periodic activation confirmed grid-like behavior.
  2. Positive correlation between directional coherence and spatial memory performance.
  3. Faster navigation speeds resulted in stronger periodic signals.
46
Q

How does speed influence the grid-like response in humans?

A

Faster navigation speeds resulted in stronger periodic signals, suggesting speed strengthens grid-like responses.

47
Q

What was the hypothesis of Constantinescu et al. (2016) regarding grid-like coding in humans?

A

They hypothesized that abstract conceptual knowledge is encoded using grids, predicting 60° periodic signals in abstract conceptual tasks.

48
Q

What cognitive map space did Constantinescu et al. (2016) investigate?

A

They studied grid cells encoding an abstract space of neck and leg lengths of birds.

49
Q

What was analyzed during each trial of the abstract conceptual task?

A

The analysis focused on the vector in 2D space, representing changes in the bird’s layout, rather than its specific location.

50
Q

What type of grid-like coding result was observed in the abstract task?

A

Hexagonal coding in the entorhinal cortex (EC) and other regions, such as the default mode network.

51
Q

How were 60° periodic signals related to trajectories and bird layout?

A

Stimuli aligned with grid axes showed stronger periodic activation, linking abstract dimensions (neck and leg lengths) to spatial grid coding.

52
Q

What is not well understood about transforming sensory information into world-centered representations?

A

How the brain processes localized sensory information (Gaussian basis) into abstract, world-centered representations (Fourier basis).

53
Q

What does the brain need to integrate to achieve world-centered representations?

A

Navigational cues such as self-motion, scene perception, and navigational cues.

54
Q

What does the diagram of different connected brain areas suggest about navigational processing?

A

Different brain regions, including the entorhinal cortex and association areas, converge to integrate navigational information.

55
Q

How are grid-like coding concepts being used in AI research?

A

Researchers are recreating grid-like coding patterns in neural network models to mimic brain navigation systems.

56
Q

what is the contast of a stimulus

A

the difference between the max and min value relative to its mean value

57
Q

increasing the contrast of the image

A
  • increases the strength of the response
  • done by amplifying the difference between the maximum and minimum pixel values relative to their mean
  • Mathematically, this means that every pixel value in the image is multiplied by a factor greater than 1
  • i.e., we can vary the contrast of a stimulus by multiplying the image itself
58
Q

mosers

A

found that many EC neurons have repetitive place fields. they activate whenever the animal is in any of a whole set of locations