Neuron models Flashcards
(25 cards)
What are HH famous for
discoveries concerning the ionic mechanicsms involved in excitation and inhibition in the peripheral and central portions of the nerve cell membrane
Of what two parts did the action potential consist according to HH
- a rapid inward current carried by Na ions
- a slowly activating outward current carried by K ions
How did they model the membrane current changes
in terms of pores/channels that were either open or closed. Then they could generate prediction for the probability of channels being open
Circuit model of HH
capacitor: represents the lipid bilayer storing charge
ion channels: resistors that allow ion flow when open
Batteries: represent equilibrium potentials driving ion currents
Current source: injected current from experiments
What is the LIF
capturing essential dynamics like integration of input and threshold-based spiking without the complexity of detailed ion channels.
Action potentials are described as events
Two components of LIF
equation that describes changes in the membrane potential (linear diff eq)
a mechanism to generate spikes (threshold for spike firing)
what is the PSP
post synaptic potential. Part of LIF. casued by the arrival of a psike from neuron j at the synapse of neuron i.
can be postivie: excitatory PSP
can be negative: inhibitory PSP
what is firing time in LIF
a neuron integrates incoing signals over time, gradually building up its internal voltage. once this voltage reaches a specific threshold, the neuron decides to fire, sending an electrical spike to communicate with other neurons. The exact moment when the voltage crosses this threshold is called the firing time, marking the instant the neuron becomes active and transmits information
what is the leaky term
without input, the potential decays back toward the resting level, as there is always a small amount fo current leaks across the membrane. in the formula RI and I are 0.
plasticity
how neurons adapt their responses over time based on activity patterns
basic hebb rule
when two neurons are repeatedly actvated at the same time, the synapse connectin them becomes stronger “neurons that fire together, wire together”. this makes it easier for neuron A to activate neuron B in the future
- the weights will increase
fundamental mechanism for learning and memory, as it helps reinfroce neural pathways that are frequently used
problems with the hebb rule
- positive feedback loop causes instablility: as synapse weights increase they can continue to grow exponentially, leading to unstable neural networks where certain connections become overly dominant
- no competition among synapses: normally they must compete for limited recourses, now that they dont all sysnapses could strengthen uniformly
- cannot model LTD since firing rates are always positive –> no forgetting
STDP
spike timing dependent plasticity: a refined model of synaptic plasticity that extends hebbs rule by incorporating the precise timing of spikes betwen neurons. the change in synaptic weight depends on the temporal order and interval between the presynaptic and postsynaptic spikes.
- if the presynaptic neuron fires just before the postynaptic neuron, the synapse is strenthened - long term potential LTP
- if the presynaptic neuron fires fter the postsynaptic neuron, the synapse is weakened - long term depression (LTD)
generalized linear models
model how inputs are transformed into neural responses
data driven insights
bridging models with experimental data
predicting spike responses from stimuli and spike history
poisson distributions
desribe rare events
E.g if you’re modeling a neurons spike count in response to a visual stimulus, a poisson GLM lets you relate stimulus features to the neurons firing rate in a principled, statistically robust way
What are ANNs
Artificial neural networks are computational models inspired by the structure and function of the human brain, they are made up of layers of interconnected neurons, which process information and learn patters from data. ANN learn by adjusting the weights based on errors in their predicitions using algorithms like backpropagation and gradient descent. this process continues iteratively to minimize the difference between the networks output an the actual target.
SO it allows us to form hupotheses about how the brain solves problems, with much better control, flexibility and overview than in real experimentse
recurrent neural network
looped connectivity allows for feedback interactions between neurons e.g. hopfield
feedforward neural network
information moves in only one direction
associative memory
the ability to perform pattern recogntition cued by partial and noisy information. problem: struggling to recreate that ability
action potential
a depolarizing stimulus opens Na channels, causing more Na to enter the cell (positive feedback)
this depolarization triggers K channels to open and Na channels to inactivate, leading ot repolarization
the delayed closiing of K channels causes a brief hyperpolarization
the system returns to the resting potential due to the leak current and channel resetting
how can we model synaptic plasticity in the nervous system
using unit (neuron/area) activity and connection weights over time as variables, and with a specific learning rule that describes how connection weights change. by manipulating the activity over time variables, changes in connection weight variables will be constructed via the learning rule
why do we want biological plausibility? what advantages can it have to accept less biological plausibility?
its at odds with computational simplicity.
what is one important neural phenomenon that the HH model can capture but the LIF cannot
the HH model is good if you want to for instance investigate the refractory period of a single neuron, or in general if you are interested in details of the stages of action potentials. The LIF simplifies these stages with only one threshold.
how does LIF simplify neuronal dynamics
it calculates the PSP by using membrane potential measurements. it has less detailed biophysical componentsm but positive or negative PSP indicates inhibitory or exitatory post-synaptic activity.
the role of the leaky term is to reset the membrane potential to resting potential when there is no external input (this is hyperpolarization of neurons). the spike threshold is a constant that when exceeded drives the membrane potential to reach action potential, which models depolarization
SO membrame potential increases with input, leaks back to resting when theres none. when potential hits threshold a spike is triggered and then reset.
It doesnt model ion channels or actual spike shapes
