Cognitive Neuroscience

1
Cognitive Neuroscience

• How do we connect cognitive processes in the mind
  with physical processes in the brain?

2
Neural plausibility

• If a cognitive process cannot be implemented by
  neurons, it cannot take place in the brain.
• How do we establish whether a process can be
  implemented by neurons?

3
Two Goals

• Localization of function - At the macro level,
  Cognitive Neuroscience tries to examine where in
  the brain various cognitive operations take
  place.
• Neural computation - At the micro level, we try
  to understand how the brain performs various
  operations.

4
Localization of function

• Post-mortem lesion studies - Find someone who
  displays an interesting cognitive deficit. When
  they die, study their brain for where the damaged
  tissue was. (Phineas Gage, Brocas Wernickes
  areas)
• Human-lesion studies - These days, we can take
  pictures of the brain while its still in the
  skull (CAT, MRI) an determine where the lesions
  are while someone is still alive. (Prosopagnosia,
  optic aphasia)

5
More localization

• Animal lesion studies - Human lesions are messy
  and uncontrolled. No two people ever have the
  exact same lesion. With animals, we can control
  the characteristics of the lesions. (Area MT)
• Single-cell recording - Also with animals, we can
  attach electrodes to neurons and measure the
  firing pattern of individual neurons. (Feature
  detectors in area V1)

6
But wait! Theres more

• Brain imaging - Modern technology provides us the
  ability to, in very broad strokes, examine what
  the brain is doing while a person is actively
  performing a cognitive task. (Face recognition,
  spatial processing)
• Brain stimulation - People are kept awake during
  brain surgery. Use the opportunity. (Motor and
  somatosensory homunculi)

7
Neural computation

• Physical modeling - Understand the properties of
  neurons, how they share information and what-not,
  and try to understand how these properties can
  lead to complex computations. (opponent
  processes, how feature detectors are calculated).
• Computational modeling - Neural networks are
  computer models of how groups of neurons behave.
  Use these models to try and better understand
  cognitive processing in the brain.

8
Neurons

• Neurons are the basic unit of the brain.
• Any information processed in the mind is
  processed by neurons.
• How can simple neurons generate complex behaviors?

9
Neural computation

• Three parts
• Dendrite Receives input from other neurons
• Cell Body Processes input
• Axon Decides and generates output

10
(No Transcript)
11
Action Potential

• The signal a neuron generates down its axon is
  called an action potential.
• All action potentials are the same magnitude
  (strength).
• We determine how excited a neuron is by its
  firing rate - how many action potentials per
  second it generates.

12
Neurotransmitter

• Neurons communicate by sending chemical messages
  called neurotransmitters to other neurons.
• These neurotransmitters travel from axon to
  either the dendrite or the cell body across the
  synapse.
• Where a synapse is depends on what the connection
  type is
• Excitatory Axon to dendrite
• Inhibitory Axon to cell body

13
Neurons as information processors

• The interesting question, from a cognitive
  perspective, is how neurons can process
  information. This leads to two questions
• How can neurons be used to perform computations?
• How can neurons store information?

14
Neural Networks

• A group of neurons acting towards a common
  purpose, such as a computation or process, is a
  neural network.
• For performing computations, we speak of
  activation that flows through the network.
• Activation represents information
• Flow represents the processing of that
  information
• Different types of information will cause
  different patterns of activation as input to the
  network
• Different patterns of light and dark on the
  retina activate different visual receptors.

15
How neurons learn

• How activation flows through a network is
  dictated by the connections in the network
• Where they are
• Excitatory vs. inhibitory
• Neural networks learn by creating, eliminating,
  or modifying connections between neurons.

16
Information storage in NNs

• So where is knowledge stored in a neural
  network?

The connections!