CH 4 (2) - Neurons and the Brain Flashcards
What does the cortical homunculus represent?
Illustrates how much space the sensory representations of different parts of the body occupy
in the cerebral cortex.
What does the Basal Ganglia do?
The basal ganglia are a group of nuclei (clusters of neurons) located beneath the cerebral cortex. The function of the basal ganglia is mainly related to the motor system where they are engaged in the planning, selection, execution and learning of actions.
What is the hippocampus?
sea horse like shape. it is involved in the formation of episodic long-term memories.
What is the function of the Amygdala?
group of nuclei related to the analysis of the emotional or motivational significance of stimuli.
What are the hippocampus and the amygdala parts of?
The limbic system
What three properties make the human brain stand out among other species?
- Neuron morphology
- Electrical properties of neurons
- Sequence processing
Neuron morphology of human brain
The pyramidal neurons in the human cortex have the most elaborate and spine-rich dendritic trees.
Electrical properties of neurons in the human brain
the pyramidal neurons in the human cortex have a very low membrane capacitance (storing of electric energy), which significantly enhances signal transmission.
Sequence processing in the human brain
Compared to other species, humans have superior capabilities of learning sequences of stimuli
When were neurons discovered?
in 19th century (1870s) by Golgi, Ramón and Cajal
When was the birth of artificial neural networks?
1943 when a computational model of the nervous system was developed.
- Every cell has only two states ( (firing and resting)
- A cell only fires if its input crosses a defined threshold
-> Starting point of modern AI
What is the perceptron and when was it developed?
1957: single neuron that can learn to classify input patterns.
- the neuron learns from falsely classifying samples by adapting its synaptic weights
When was the AI Winter?
- The perceptron only works for simple data sets that are linearly separable (Xor problem cannot be solved).
Why has the AI winter ended?
With the development of new neural network architectures and the back-propagation algorithm, which has become the basis of all modern neural networks architecture.
What are biological neurons?
Cells with a highly intricate structure and complex electrophysiological dynamics
What are analog neuron models?
real-valued (real numbers) output that can be interpreted as the neuron’s firing rate
( artificial neural network in which an electrically adjustable material is used to emulate the function of a neural synapse.)
spiking neural network
emits impulses that correspond to the spikes emitted by biological neurons
Analog neuron model forumla
output = n(x) = A(w1x1 + … + wnxn + b)
Common activation functions
Binary step function Logistic function tanh Rectified Linear Unit (ReLu) Exponential Linear Unit (ELU) Gaussian
What is information processing in biological neurons largely driven by?
ionic currents that accumulate in the soma (cell body)
How can the flow of the currents in the soma be modeled?
functionally equivalent electrical circuits
What is the Leaky Integrate-and-Fire Model?
first spiking neuron model that is still of relevance today. Comprises of a electric circuit composed of resistors and capacitors. It may be driven by a voltage or current source and these will produce different responses.
Does the LIF neuron capture cell morphology?
No it is a point neuron model?
More detailed spiking neuron models are derived from the
Hodgkin-Huxley model
What are the 5 levels of modeling?
- Detailed compartmental models
- Reduced compartmental models
- Single-compartment models
- Cascade models
- Block-box models
What is the detailed compartmental model?
- captures the detailed neuron morphology based on anatomical reconstructions. The continuous geometrical structure of the neuron is discretised into compartments, each of which can be modeled as a point neuron.
What is the Reduced compartmental model?
comprised of only a few dendritic compartments to decrease computational complexity and to increase analytical tractability.
What is the single-compartment model?
ignores the morphology of the neuron. Detailed spiking point neuron models capture the flow of ion currents and spike generation.
What is the cascade model?
purely functional model that only represents abstract computations but no biological detail.
What is the black-box model?
Does not model any computational function at all but considers the neuron as an input/output system with defined statistical behavior.
Are standard methods in neuroscience sufficient the way they are applied now?
Not really. If they are used on microprocessors they do not lead to a meaningful understanding of the processor in terms of information processing hierarchy. However, a hypothesis would be reasonably easy to test.
What is whole brain modeling?
Executable hypothesis of the architecture and the functioning of the brain. Whole brain models integrate a large number of simulated neuron models in a coherent neural network model that captures to connectivity of biological brains to some degree.
Many whole brain models are simulations of
the thalamocortical system
Simplified engineering brain models are defined manually based on
rules and statistics identified in datasets