Lecture 07 - Neural networks and biology Flashcards
Neural networks and biology
What is this part called?
An apical dendrite
Neural networks and biology
What are these called?
Basal dendrites
Neural networks and biology
What are basal dendrites?
A basal dendrite is a dendrite that emerges from the base of a pyramidal cell that receives information from nearby neurons and passes it to the soma, or cell body.
Neural networks and biology
What are apical dendrites?
An apical dendrite is a dendrite that emerges from the apex of a pyramidal cell.
Neural networks and biology
What is this part called?
Soma
Neural networks and biology
What is this part called?
Axon
Neural networks and biology
Where is the apical dendrite located?
See image
Neural networks and biology
Where is the soma located?
See image
Neural networks and biolog
Where is the axon located?
See image
Neural networks and biology
Where is the basal dendrite located?
See image
Neural networks and biology
What are McCulloch and Pitts known for?
Proposing the first computational model of a neuron.
Neural networks and biology
Who proposed the first computational model of a neuron?
McCulloch and Pitts
Neural networks and biology
When did McCulloch and Pitts first suggest their neuron model?
1943?
Neural networks and biology
Describe what’s special about the inputs and outputs of McCulloch-Pitts neurons
They’re boolean values.
Neural networks and biology
What’s the activation function of the McCulloch-Pitts neuron?
The step function
Neural networks and biology
Are the connections in the McCulloch-Pitts model weighted or unweighted?
Unweighted.
Neural networks and biology
Who proposed the perceptron?
Frank Rosenblatt
Neural networks and biology
When was the perceptron introduced?
1958
Neural networks and biology
What neural model was introduced in 1943?
The McCulloch-Pitts neuron model
Neural networks and biology
What neural model was introduced in 1958?
The perceptron
Neural networks and biology
Does the perceptron use weighted or unweighted connections?
Weighted
Neural networks and biology
What kind of input does the perceptron accept?
Real valued numbers.
Neural networks and biology
Which network only accepts boolean inputs? (Perceptron or McCulloch-Pitts model)
The McCulloch-Pitts model
Neural networks and biology
Which activation function is used in the perceptron?
Trick question - you can choose.
Neural networks and biology
What is the output type of the perceptron?
Boolean outputs.
Maybe that’s just Rosenblatt’s initial model from 1958?
Neural networks and biology
Which model is this?
Rosenblatt’s perceptron
Neural networks and biology
Which model is this?
A McCulloch-Pitts neuron
Neural networks and biology
Why do we model neurons? (2)
1) Understanding
2) Inspiration
Neural networks and biology
What is a synapse?
A connection point between two neurons.
Neural networks and biology
What is a dendrite?
The input channel to a neuron.
Neural networks and biology
What is an axon?
Output channels from a neuron.
Neural networks and biology
What is transferred through an axon?
Neurotransmitters.
Neural networks and biology
What types of ions determine the cell’s potential?
Sodium (Na+) and potassium (K+).
Neural networks and biology
What are vesicles?
Small fluid-filled containers with neurotransmitters.
Neural networks and biology
What are the small fluid-filled containers with neurotransmitters called?
Vesicles.
Neural networks and biology
What are vesicles filled with?
Neurotransmitters.
Neural networks and biology
What types of synapses exist? (2)
Excitatory and inhibitory.
Neural networks and biology
What part of the neuron can be either excitatory or inhibitory?
The synapse.
Neural networks and biology
What do weights in an ANN represent (superficially)?
Whether a synapse is excitatory or inhibitory.
Neural networks and biology
What happens when a signal comes in from a synapse and arrives at the neuron?
The neuron “integrates” (=sums) it.
(Metaphor of charging a capacitor.)
Neural networks and biology
What does the feedforward excitation neuron connection look like?
Neural networks and biology
What kind of neuron connection is this an example of?
Feedforward excitation.
Neural networks and biology
What does the feedforward inhibition neuron connection look like?
Neural networks and biology
What kind of neuron connection is this an example of?
Feedforward inhibition
Neural networks and biology
What does the convergence/divergence neuron connection look like?
Neural networks and biology
What kind of neuron connection is this an example of?
Convergence/divergence
Neural networks and biology
What does the lateral inhibition neuron connection look like?
Neural networks and biology
What kind of neuron connection is this an example of?
Lateral inhibition
Neural networks and biology
(@image missing) What does the feedback/recurrent inhibition neuron connection look like?
Neural networks and biology
What kind of neuron connection is this an example of?
feedback/recurrent inhibition
Neural networks and biology
What does the feedback/recurrent excitation neuron connection look like?
Neural networks and biology
What kind of neuron connection is this an example of?
feedback/recurrent excitation
Neural networks and biology
What is synaptic plasticity?
Synaptic plasticity is the change in the connection strength that occurs at synapses, meaning how active or inactive they are.
Neural networks and biology
Who coined the term synaptic plasticity?
Donald Hebb
Neural networks and biology
When did Donald Hebb coin the term synaptic plasticity?
1949
Neural networks and biology
Synapses can change the strength of the signals they send. What is this called?
Synaptic plasticity.
Neural networks and biology
What is long-term potentiation? (abbrev. LTP)
Long-term potentiation (LTP) is a persistent increase in synaptic strength following high-frequency stimulation of a chemical synapse.
Neural networks and biology
What is it called when synaptic strength increases after high-frequency stimulation of a chemical synapse?
Long-term potentiation (LTP)
Neural networks and biology
Who discovered long-term potentiation?
Terje Lømo
Neural networks and biology
When did Terje Lømo discover long-term potentiation?
1966
Neural networks and biology
Are synapses chemical or electrical?
Both
Neural networks and biology
Is chemical or electrical synapses more common?
Chemical
Neural networks and biology
Describe Hebbian plasticity
If a pre-synaptic neuron fires shortly before a post-synaptic neuron, their connection is strengthened.
Neural networks and biology
Describe the topology of a biological neural network.
Sparse and complex. Not dense, not regular, yet not random.
Neural networks and biology
Describe what’s done because biological connections are metabolically costly to maintain.
They are pruned.
Neural networks and biology
How do neurons communicate?
Via action potentials (AP).
Neural networks and biology
How are action potentials generated?
By ions travelling across neural membranes.
Neural networks and biology
What is generated when ions travel across neural membranes?
Action potentials.
Neural networks and biology
What do conventional ANNs approximate?
Rate coding.
Neural networks and biology
Which type of network approximate rate encoding?
Conventional ANNs.
Neural networks and biology
Describe the leaky-integrate-and-fire (LIF) model. (4)
1) Spikes are sent into the neuron.
2) The neuron sums all incoming spikes.
3) The neuron “leaks” (the signal decays).
4) If the spike crosses a threshold, a spike is sent out and the neuron reset.
Neural networks and biology
What are some “common sense” coding methods for spiking neural networks (SNNs)?
1) Rate coding
2) Time-to-first-spike coding
3) Phase coding
4) Burst coding
Neural networks and biology
What is STDP short for?
Spike-time dependent plasticity
Neural networks and biology
What are the ways of training SNNs?
1) STDP learning
2) Stochastic STDP
3) ANN-SNN conversion
4) Backpropagation
Neural networks and biology
What is neuromorphic hardware?
Hardware for running SNNs.
