Supercharging Your Studying with ChatGPT & Anki

## Eye Movement Control System

- researchers are able to gauge decision-making within primates by tracking their eye movements over different targets that appear on the screen
  - We turn this into a decision-making task by allowing the primates to choose between different targets on the screen, and potentially get rewarded based off of the target that they select based off of their eye

### Superior Colliculus

- Superior colliculus contains a “motor map” of visual space
  - There are different neurons that map to different places your eyes might gaze, and each of the neurons have movement fields
  - To trigger an eye movement, where you direct your gaze from one target to another, the cells that map to the target, increase their firing rates sharply, which will evoke saccades, according to topographic organization
  - They also show anticipatory activity meaning that if you're thinking about moving your gaze, they might also start firing a little more
- Superior colliculus & multiple potential targets: during the uncertainty period the SC activity for the particular saccade scales with the number of targets, just like the expected value
  - 1 target: EV(target) = 100% \* reward magnitude
  - 2 targets: EV(target) = 50% \* reward magnitude
  - 4 targets: EV(target) = 25% \* reward magnitude
  - 8 targets: EV(target) = 12.5% \* reward magnitude
- during the uncertainty period, the SC activity for a particular saccade scales with the number of targets, just like EV, but when the target is selected, the firing rate is similar regardless of how many initial targets there are

**Resolving Decision Making**

1. Population vector representations
2. Winner-Take-All Dynamic

### Population Vectors

- Cells near the stimulation site respond to saccades in many different directions and SC neurons are generally broadly tuned to saccade direction
- So lets say we have a target for the eye movement at A, neurons B and C will still be slightly active. They each will have their preferred direction and they are kind of fighting for their direction to be selected
  - The population vector will be the vector sum of the different neurons that will fire during the target within the SC
- Consider, B we are asking for the target to be at B and end up getting something closer to in between B and D because we deactivated A
- In executing a single eye movement, all of the general neurons within the vicinity will contribute to where the gaze is directed

---

**Population Vector Algorithm:**

1. Each cell comprises a vector emanating from the origin and aligned along its preferred direction.
2. The length of each vector is dependent on the neurons firing rate
3. The reach angle is the vector sum of all of the neuron’s vectors

---

- the great part about the population vectors is the idea that we have redundancy within the neurons so that if some neurons do not work, it will not destroy the entire system
  - also, the noise in individual neurons gets averaged out
- Incoherent or inconsistent representations can lead to maladaptive behaviors due to
  vector averaging

### Winner Takes all Networks

- When a neuron is excited, it will excite other neurons that are nearby and have a similar preferred direction
- The inhibition portion allows for one corehent thing will get decided on. This occurs when one neuron will get ahead of the others.

[
    {
        "front": "How do researchers gauge decision-making in primates?",
        "back": "By tracking their eye movements over different targets on the screen."
    },
    {
        "front": "How do primates participate in a decision-making task in eye movement studies?",
        "back": "Primates choose between different targets on the screen and may receive rewards based on their selected target."
    },
    {
        "front": "What is the role of the superior colliculus in eye movement control?",
        "back": "The superior colliculus contains a 'motor map' of visual space, with neurons mapping to different places the eyes might gaze."
    },
    {
        "front": "What happens to neurons in the superior colliculus when you direct your gaze from one target to another?",
        "back": "Neurons that map to the target increase their firing rates sharply, triggering saccades."
    },
    {
        "front": "What is anticipatory activity in the superior colliculus?",
        "back": "It refers to neurons increasing their firing rates when you're thinking about moving your gaze, even before the movement occurs."
    },
    {
        "front": "How does superior colliculus activity scale with the number of potential targets during the uncertainty period?",
        "back": "SC activity for a particular saccade scales with the expected value of the reward, decreasing as the number of targets increases."
    },
    {
        "front": "What happens to the firing rate in the superior colliculus when a target is selected, regardless of the number of initial targets?",
        "back": "The firing rate becomes similar, regardless of how many initial targets there were."
    },
    {
        "front": "What are the two key processes involved in resolving decision-making in eye movement control?",
        "back": "Population vector representations and Winner-Take-All dynamics."
    },
    {
        "front": "How do superior colliculus neurons respond to saccades in different directions?",
        "back": "Neurons are broadly tuned to saccade direction and respond to multiple directions, contributing to population vector formation."
    },
    {
        "front": "What is a population vector in the context of eye movement control?",
        "back": "A vector sum of different neurons firing in the superior colliculus, representing the direction of the eye movement."
    },
    {
        "front": "How does deactivating certain neurons affect the population vector and eye movement?",
        "back": "Deactivating certain neurons can shift the eye movement target, as the population vector is the sum of all active neurons' contributions."
    },
    {
        "front": "What ensures redundancy and resilience in the population vector mechanism?",
        "back": "Redundancy in neurons ensures that even if some neurons fail, the overall system continues to function effectively."
    },
    {
        "front": "How does the population vector mechanism help mitigate noise in individual neurons?",
        "back": "Noise gets averaged out across the population, ensuring more accurate movement direction."
    },
    {
        "front": "What can happen if there are incoherent or inconsistent population vector representations?",
        "back": "Incoherent or inconsistent representations can lead to maladaptive behaviors due to vector averaging."
    },
    {
        "front": "What happens when a neuron excites nearby neurons with similar preferred directions in a Winner-Takes-All network?",
        "back": "The excited neuron stimulates nearby neurons with similar directions, eventually leading to one coherent decision."
    },
    {
        "front": "How does inhibition contribute to the Winner-Takes-All dynamic?",
        "back": "Inhibition allows one dominant neuron to get ahead of others, leading to a clear decision."
    }
]

# anki_script.py

import json
import urllib.request

deck_name = "sample_deck"
flashcards_path = "flashcards.json"

def request(action, **params):
    return {'action': action, 'params': params, 'version': 6}

# Add Card and Catch Errors
def invoke(action, **params):
    requestJson = json.dumps(request(action, **params)).encode('utf-8')
    response = json.load(urllib.request.urlopen(urllib.request.Request('http://127.0.0.1:8765', requestJson)))
    if len(response) != 2:
        raise Exception('response has an unexpected number of fields')
    if 'error' not in response:
        raise Exception('response is missing required error field')
    if 'result' not in response:
        raise Exception('response is missing required result field')
    if response['error'] is not None:
        raise Exception(response['error'])
    return response['result']

# Load Flashcards from Json
with open(flashcards_path, 'r') as file:
        flashcards = json.load(file)
        num_flashcards = 0

        for flashcard in flashcards:
            front_content = flashcard['front']
            back_content = flashcard['back']

            # Create Flashcards
            num_flashcards += 1
            result = invoke(action='addNote', note={"deckName": deck_name, "modelName": "Basic", "fields": {"Front": front_content, "Back": back_content}})
            print('✅ Successfully Created Card {}: {}'.format(num_flashcards, result))

        print('\n ✅ Created {} flashcards.'.format(len(flashcards)))

// flashcards.json

[
  {
    "front": "How do researchers gauge decision-making in primates?",
    "back": "By tracking their eye movements over different targets on the screen."
  }
]

$ python3 anki.py