Task 4 - Connectionism and BCI's Flashcards
Connectionist network
Simple representations consisting only of units and links
Units: neuronlike components that have a degree of activation corresponding to firing rate of neuron
Links: Connections beteen units, can be exhibitary or inhibitory and can be one-way or both ways
Two classes of connectionist models
Class 1 - Concerned with local representations (specifiable components)
Class 2 - Concerned with distributed representations, where meaning is distributed over multiple components
Hebbian learning
Suggests that if two neurons are activated at the same time, the connection between them should be strengthened (what fires together wires together)
Pattern association
The task of learning to associate one stimulus with another
Unsupervised learning
- Type of machine learning where the model is trained on data without labeled outputs or explicit supervision.
- System tries to identify patterns, relationships, or structures in input data on its own, without being provided specific answers or categories.
Competitive learning
- Type of unsupervised learning
- Output units compete with each other to determine which has the largest response to a pattern. That unit is then responsible
Graceful degradation
- Hallmark of distributed representation
- Refers to the ability of a cognitive system to be robust to the loss of parts of its system
BCI (Brain-Computer Interface)
A system that enables direct communication between the brain and an external device, bypassing traditional output mechanisms like muscles or speech, using:
1. Signal Aquisition
2. Signal Processing
3. Output device
Clinical classification of BCIs
Assistive BCIs: Aim to substitute lost functions
Rehabilitative BCIs: Aim to facilitate the restoration of brain function
5 main types of brain activity measured with invasive BCIs
- Local field potentials
- Single-unit activity
- Multi-unit activity
- Electrocorticographic oscillations
- Calcium channel permeability
7 types of signal measured by noninvasive BCIs
- Slow cortical potentials
- Sensorimotor rhythms
- P300 event-related potential (EEG)
- Steady-state visual evoked potentials (EEG)
- ERNPs (EEG)
- BOLD signal (fMRI)
- Cerebral oxygenation changes (fNIRS)
Autoassociation
Type of memory process where a neural network strengthens connections between similar or previously encountered inputs, enabling pattern completion and recall from partial information.
Delta rule
Learning rule in neural networks that adjusts connection weights based on the difference (error) between the predicted and actual output
