L12-PatternAssociator1 (1)

Cognitive Neuroscience: Overview

• Cognitive Science: Interdisciplinary field studying the mind and its processes, including thinking, learning, memory, and language.

• Modeling Approaches: Includes computational modeling to simulate cognitive processes.

• Core Topics:

  • Decision Making

  • Learning and Memory

  • Perception

  • Language

Readings for Lectures

Today's Lecture

• Lecture slides provided.

• McLeod, Plunkett, & Rolls (1998). Intro to Connectionist Modeling of Cognitive Processes, pp. 30-35, 48-60.

• Cacioppo & Freberg (2012). Discovering Psychology, pp. 442-443.

Next Lecture

• Lecture slides provided.

• McLeod, Plunkett, & Rolls (1998). Intro to Connectionist Modeling of Cognitive Processes, pp. 30-35, 54-64, 72-74.

• Optional Reading: McClelland (2000). Connectionist Models of Memory.

Talk Outline

• Neural Representations

• Pattern Associator (Part 1)

• Synaptic Plasticity

Overall Theoretical Framework

• Mental States: Represented as patterns of activation in a distributed memory system.

• Mental Processing: Transformation of input patterns to output patterns.

• Knowledge Representation: Found in the synaptic weights.

• Memory Traces: Changes in synaptic weights that facilitate retrieval.

• Reference: McClelland & Rumelhart (1985).

Neural Representations and Processing

• The brain transforms external entities into patterns of activation across neuron populations.

• Information processing governed by:

  • Synaptic weights affecting transformation efficiency.

  • Representational Vehicles: Neural coded inputs and outputs.

  • Representational Content: Information encoded by neural activation patterns.

Examples of Code Messages

• Lists of codes representing actions needed by military units in different quadrants (A, B, C, D) using binary-based coding.

  • E.g., tanks in quadrant A need reinforcements.

Localist Neural Representations

• Each stimulus feature (size, color, shape) coded by distinct neuron pools.

• Individual values assigned to dedicated "detector" neurons (e.g., color detectors).

• Facilitates linguistic description of stimuli.

Distributed Neural Representations

• Each item represented as patterns of activity over multiple neurons.

• Same neuron contributes to multiple item representations; activation pattern determines the represented item.

• Examples: Patterns for colors like red, blue, and green utilize overlapping neural populations.

Input-Output Mapping

• Definition: Mapping that illustrates correspondence between input and output patterns (e.g., phone numbers to names).

• Mathematical equivalent termed as functions (y = f(x)).

Pattern Associator Network

• Explains the structure that maps input patterns (x) to output patterns (y).

• Can visualize using example activations.

Simulated Neuron Functionality

• The network includes inputs and connection weights similar to synaptic efficacy.

• Net input computed as weighted sum of inputs.

• Activation thresholds determine whether outputs occur.

Synaptic Plasticity

• Definition: Changes in synaptic strength influenced by experience; can be either strengthening or weakening.

  • Key Influences: Brain's ability to learn and adapt—central to learning mechanisms in the brain.

• Hebbian Learning: Increases efficiency of synaptic connections between neurons that activate simultaneously—"cells that fire together wire together".

Molecular Mechanisms of Synaptic Plasticity

Receptors Involved

• AMPA Receptors: Open during glutamate binding, facilitating Na+ entry.

• NMDA Receptors: Opens when glutamate binds AND the neuron membrane is depolarized, allowing Ca2+ influx.

Scenarios of Synaptic Activity

1. No Postsynaptic Spike: Less activation, no learning.

2. Postsynaptic Spike: Activation results in synaptic strengthening, facilitating learning.

Two Mechanisms of Synaptic Plasticity

1. Early Synaptic Plasticity: Fast-acting, temporary, no protein synthesis needed.

2. Synaptic Consolidation: More permanent, requires protein synthesis.

Optional Material

• Later slides contain additional information not critical for immediate study but provide further insights into neuroscience topics.