Q.1 Which differential equation best describes the classic Hodgkin‑Huxley model of the membrane potential?
C_m \( \frac{dV}{dt} = I_{ext} - g_{Na}m^3h(V-E_{Na}) - g_K n^4 (V-E_K) - g_L (V-E_L) \)
C_m \( \frac{dV}{dt} = I_{ext} - g_{L}(V-E_L) \)
\( \tau \frac{dV}{dt} = -(V-V_{rest}) + R I_{ext} \)
V(t) = V_{rest} + R_m I_{ext} (1 - e^{-t/\tau_m})
Explanation - The Hodgkin‑Huxley model uses four conductance terms (Na, K, leak) with gating variables m, h, n to capture the voltage‑dependent dynamics of the squid giant axon.
Correct answer is: C_m \( \frac{dV}{dt} = I_{ext} - g_{Na}m^3h(V-E_{Na}) - g_K n^4 (V-E_K) - g_L (V-E_L) \)
Q.2 In a leaky integrate‑and‑fire neuron, the membrane time constant τ_m is defined as:
τ_m = R_m C_m
τ_m = \frac{C_m}{R_m}
τ_m = R_m + C_m
τ_m = \frac{R_m}{C_m}
Explanation - The membrane time constant is the product of membrane resistance and capacitance, determining how fast the voltage decays toward rest.
Correct answer is: τ_m = R_m C_m
Q.3 Which of the following coding schemes assumes that information is carried by the precise timing of spikes rather than their rate?
Rate coding
Temporal coding
Population coding
Binary coding
Explanation - Temporal coding emphasizes the exact timing of individual spikes (e.g., latency or phase) as the primary information carrier.
Correct answer is: Temporal coding
Q.4 Spike‑Timing‑Dependent Plasticity (STDP) typically leads to Long‑Term Potentiation (LTP) when:
A presynaptic spike follows a postsynaptic spike within ~20 ms
A presynaptic spike precedes a postsynaptic spike within ~20 ms
Both spikes occur simultaneously
The postsynaptic neuron is hyperpolarized
Explanation - In classic STDP, presynaptic spikes that arrive shortly before postsynaptic spikes strengthen the synapse (LTP), while the reverse order induces depression (LTD).
Correct answer is: A presynaptic spike precedes a postsynaptic spike within ~20 ms
Q.5 Which algorithm is most commonly used to separate independent source signals from mixed EEG recordings?
Principal Component Analysis (PCA)
Independent Component Analysis (ICA)
K‑means clustering
Fourier Transform
Explanation - ICA assumes statistical independence of source signals and can isolate artifacts (e.g., eye blinks) from neural activity in EEG data.
Correct answer is: Independent Component Analysis (ICA)
Q.6 The term "receptive field" of a neuron in visual cortex most directly refers to:
The set of synaptic inputs the neuron receives
The region of visual space that modulates the neuron’s firing
The range of membrane potentials the neuron can attain
The temporal pattern of spikes generated
Explanation - A receptive field defines the specific area of the visual field where a stimulus influences the neuron's activity.
Correct answer is: The region of visual space that modulates the neuron’s firing
Q.7 Which of the following is a hallmark of a chaotic neural network dynamics?
Periodic firing with fixed frequency
Exponential divergence of nearby trajectories
Linear relationship between input and output
Stable fixed point attractor
Explanation - Chaos is characterized by sensitive dependence on initial conditions, leading to exponential divergence of close state trajectories.
Correct answer is: Exponential divergence of nearby trajectories
Q.8 In the context of computational models, the term "synaptic weight" usually denotes:
The physical size of the synapse
The conductance change induced by a presynaptic spike
The membrane capacitance of the postsynaptic neuron
The refractory period duration
Explanation - Synaptic weight quantifies how much a presynaptic event changes the postsynaptic conductance, influencing the effect on membrane potential.
Correct answer is: The conductance change induced by a presynaptic spike
Q.9 Which of the following best describes a "population vector" used in motor cortical decoding?
A vector of membrane potentials from a single neuron
A weighted sum of preferred directions of many neurons
A binary code representing spike presence
A Fourier transform of firing rates
Explanation - Population vectors combine each neuron's preferred movement direction weighted by its firing rate to estimate the intended movement direction.
Correct answer is: A weighted sum of preferred directions of many neurons
Q.10 Which of the following is NOT a typical assumption of the Wilson‑Cowan model of neural populations?
Excitatory and inhibitory populations interact through average firing rates
Each neuron fires deterministically according to a fixed threshold
Synaptic interactions are described by sigmoidal activation functions
The model operates on a continuous time scale
Explanation - The Wilson‑Cowan model abstracts away individual spikes, using average rates; it does not model deterministic threshold spiking of individual neurons.
Correct answer is: Each neuron fires deterministically according to a fixed threshold
Q.11 In a conductance‑based synapse model, the postsynaptic current I_syn(t) is given by:
I_syn = g_syn (V - E_rev)
I_syn = g_syn V
I_syn = C_m \frac{dV}{dt}
I_syn = \frac{V}{R_m}
Explanation - Conductance‑based models calculate current as the product of synaptic conductance and the driving force (V – reversal potential).
Correct answer is: I_syn = g_syn (V - E_rev)
Q.12 Which metric quantifies the similarity between two spike trains by counting coincident spikes within a temporal window?
Pearson correlation coefficient
Victor‑Purpura distance
Jensen‑Shannon divergence
Spike‑time tiling coefficient (STTC)
Explanation - STTC measures the proportion of spikes that fall within a defined temporal tolerance, providing a robust similarity index for spike trains.
Correct answer is: Spike‑time tiling coefficient (STTC)
Q.13 Which of the following is a primary advantage of using the Izhikevich neuron model over the Hodgkin‑Huxley model?
It reproduces a wide variety of spiking patterns with very few parameters
It includes detailed ion‑channel kinetics
It models dendritic morphology explicitly
It requires solving partial differential equations
Explanation - The Izhikevich model combines biological plausibility with computational efficiency, capturing many firing regimes using only two differential equations.
Correct answer is: It reproduces a wide variety of spiking patterns with very few parameters
Q.14 In fMRI BOLD signal analysis, the hemodynamic response function (HRF) is typically modeled as:
A delta function
A gamma‑variate function
A sinusoidal wave
An exponential decay only
Explanation - The HRF is commonly approximated by a gamma‑variate function to capture its initial rise and later undershoot in the BOLD response.
Correct answer is: A gamma‑variate function
Q.15 Which of the following statements about the 'balance of excitation and inhibition' (E/I balance) in cortical circuits is true?
E/I balance only occurs during sleep
Disruption of E/I balance can lead to epilepsy
Inhibition always dominates excitation in the cortex
Excitation and inhibition are independent of each other
Explanation - A proper balance prevents runaway excitation; its disruption is linked to hyperexcitability disorders such as epilepsy.
Correct answer is: Disruption of E/I balance can lead to epilepsy
Q.16 The Nyquist theorem states that to avoid aliasing, a signal must be sampled at least at:
Twice the highest frequency present in the signal
The same rate as the highest frequency
Half the highest frequency
Four times the highest frequency
Explanation - Sampling at ≥2× the maximum signal frequency ensures the original signal can be perfectly reconstructed.
Correct answer is: Twice the highest frequency present in the signal
Q.17 Which of the following best describes the term "neuromorphic engineering"?
Designing digital computers that simulate brain activity using GPUs
Creating hardware that mimics the structure and function of neural systems
Using fMRI to map brain activity in real time
Applying machine learning to classify neural spikes
Explanation - Neuromorphic engineering builds analog or mixed‑signal circuits that emulate neuronal dynamics and synaptic plasticity.
Correct answer is: Creating hardware that mimics the structure and function of neural systems
Q.18 In a recurrent neural network (RNN) used to model a cortical column, the term "fixed point" refers to:
A state where the network output repeats every time step
A state where the network activity does not change over time
A point in parameter space where learning stops
The maximum firing rate achievable by any neuron
Explanation - A fixed point is a stable or unstable equilibrium of the network dynamics where derivatives are zero.
Correct answer is: A state where the network activity does not change over time
Q.19 Which of the following is a common method for estimating functional connectivity from multi‑unit recordings?
Cross‑correlation histogram
Mean squared error
Linear regression
Monte Carlo integration
Explanation - Cross‑correlation measures the temporal relationship between spike trains, indicating functional coupling.
Correct answer is: Cross‑correlation histogram
Q.20 The term "phase locking" in neuronal oscillations refers to:
Neurons firing at random times relative to a rhythm
Neurons firing at a constant phase of an ongoing oscillation
Neurons never firing during an oscillation
Neurons synchronizing only during sleep
Explanation - Phase locking indicates that spikes consistently occur at a particular phase of a rhythmic signal, reflecting temporal coordination.
Correct answer is: Neurons firing at a constant phase of an ongoing oscillation
Q.21 In a Hopfield network used for associative memory, the energy function is minimized when:
All neurons are silent
The network reaches a stable attractor representing a stored pattern
Synaptic weights become zero
External input is removed
Explanation - Hopfield networks converge to minima of an energy landscape that correspond to stored memory patterns.
Correct answer is: The network reaches a stable attractor representing a stored pattern
Q.22 Which of the following is a typical feature of a "bursting" neuron model?
A single spike per input pulse
Rapid succession of spikes followed by a quiescent period
Continuous, regular spiking at a fixed frequency
No spiking, only subthreshold oscillations
Explanation - Bursting neurons fire groups of spikes (bursts) separated by silent intervals, often modeled with slow and fast variables.
Correct answer is: Rapid succession of spikes followed by a quiescent period
Q.23 When fitting a computational model to electrophysiological data, the term "parameter identifiability" means:
All parameters can be estimated uniquely from the data
Parameters are interchangeable without affecting model output
The model has more parameters than data points
Parameters are fixed and cannot be changed
Explanation - Identifiability ensures that each parameter influences the model output in a distinguishable way, allowing unique estimation.
Correct answer is: All parameters can be estimated uniquely from the data
Q.24 Which of the following best describes a “neural field model”?
A model of a single neuron’s ion channels
A spatially continuous description of population activity over cortical surface
A discrete network of point neurons with random connections
A model that only simulates synaptic plasticity
Explanation - Neural field models use integro‑differential equations to capture activity propagation across continuous space.
Correct answer is: A spatially continuous description of population activity over cortical surface
Q.25 In the context of neural data analysis, the term "dimensionality reduction" most often aims to:
Increase the number of recorded neurons
Compress data while preserving the most variance or information
Eliminate noise by deleting data points
Transform data into a binary format
Explanation - Techniques like PCA, factor analysis, or t‑SNE reduce data dimensionality while retaining essential structure for interpretation.
Correct answer is: Compress data while preserving the most variance or information
Q.26 Which of the following statements about the FitzHugh‑Nagumo model is true?
It is a four‑dimensional model with detailed ion channels
It reduces the Hodgkin‑Huxley model to a two‑variable system capturing excitability
It models synaptic plasticity directly
It cannot generate action potentials
Explanation - The FitzHugh‑Nagumo model simplifies neuronal dynamics into a fast voltage variable and a slow recovery variable, preserving excitability features.
Correct answer is: It reduces the Hodgkin‑Huxley model to a two‑variable system capturing excitability
Q.27 In a conductance‑based network, excitatory synapses typically have reversal potentials near:
-70 mV
-50 mV
0 mV
+60 mV
Explanation - Excitatory (glutamatergic) synapses are often modeled with a reversal potential around 0 mV, driving depolarization.
Correct answer is: 0 mV
Q.28 The term "Hebbian learning" is often summarized by which phrase?
Neurons that fire together, wire together
Synapses weaken with activity
Neurons inhibit each other
Learning is independent of timing
Explanation - Hebbian learning postulates that coincident activity strengthens the synapse linking the active neurons.
Correct answer is: Neurons that fire together, wire together
Q.29 Which of the following is NOT a typical assumption of the leaky integrate‑and‑fire (LIF) model?
Membrane potential integrates incoming currents linearly
Spike generation is instantaneous when threshold is crossed
After a spike, the membrane potential resets to a fixed value
Synaptic conductances are modeled as nonlinear functions of voltage
Explanation - The LIF model usually uses current‑based inputs; voltage‑dependent conductance dynamics are not included.
Correct answer is: Synaptic conductances are modeled as nonlinear functions of voltage
Q.30 In a Bayesian decoder for neural population activity, the posterior probability P(s|r) is proportional to:
P(r|s) P(s)
P(s) / P(r|s)
P(r) P(s)
P(s|r) P(r|s)
Explanation - Bayes' rule states P(s|r) ∝ P(r|s) P(s), where P(r|s) is the likelihood of response r given stimulus s, and P(s) is the prior.
Correct answer is: P(r|s) P(s)
Q.31 Which of the following is a key advantage of using event‑driven (spike‑based) simulation over time‑step simulation for large neural networks?
Event‑driven simulation is always faster regardless of activity level
It reduces computation by updating only when spikes occur
It eliminates the need for random number generators
It automatically solves differential equations analytically
Explanation - Event‑driven simulators skip idle periods, processing only spike events, which can dramatically speed up sparse network simulations.
Correct answer is: It reduces computation by updating only when spikes occur
Q.32 The term "criticality" in neural systems is most closely related to:
The point at which neurons become completely silent
A balance between order and disorder where activity follows a power‑law distribution
The maximal firing rate a neuron can achieve
The threshold voltage for spike initiation
Explanation - Criticality describes a regime where the system is poised between quiescence and runaway excitation, often evidenced by avalanche size distributions.
Correct answer is: A balance between order and disorder where activity follows a power‑law distribution
Q.33 In the context of spike‑train analysis, the "inter‑spike interval (ISI) histogram" provides information about:
The average membrane potential
The distribution of time gaps between consecutive spikes
The spatial location of the neuron
The synaptic weight matrix
Explanation - ISI histograms plot the frequencies of intervals between spikes, revealing regularity, bursting, or refractoriness.
Correct answer is: The distribution of time gaps between consecutive spikes
Q.34 Which of the following is a typical use of the Kalman filter in neural engineering?
Decoding intended movement from neural firing rates in real‑time
Estimating synaptic conductance from calcium imaging
Removing motion artifacts from EEG
Generating synthetic spike trains
Explanation - Kalman filters provide optimal recursive estimation of continuous variables (e.g., limb position) from noisy neural observations.
Correct answer is: Decoding intended movement from neural firing rates in real‑time
Q.35 Which of the following is NOT a standard method for measuring neural synchrony?
Phase‑locking value (PLV)
Coherence
Cross‑entropy
Spike‑field coherence
Explanation - Cross‑entropy quantifies differences between probability distributions; it is not a direct measure of synchrony like PLV or coherence.
Correct answer is: Cross‑entropy
Q.36 In a synaptic model with short‑term depression, the available neurotransmitter resource R(t) typically follows:
R(t+Δt)=R(t)-U·R(t)·δ(t−t_spike)+ (1−R(t))/τ_rec
R(t+Δt)=R(t)+U·R(t)·δ(t−t_spike)
R(t)=constant
R(t)=exp(−t/τ_decay)
Explanation - Short‑term depression reduces the resource by a fraction U at each spike and recovers with time constant τ_rec.
Correct answer is: R(t+Δt)=R(t)-U·R(t)·δ(t−t_spike)+ (1−R(t))/τ_rec
Q.37 Which of the following best describes a "reservoir computer" as used in computational neuroscience?
A recurrent network with fixed random connections that projects input into a high‑dimensional space
A feed‑forward network trained by back‑propagation
A hardware chip that mimics ion channels
A statistical model for spike‑train generation
Explanation - Reservoir computing leverages a randomly connected dynamical system (the reservoir) whose internal states are used for downstream linear read‑out training.
Correct answer is: A recurrent network with fixed random connections that projects input into a high‑dimensional space
Q.38 The term "neuronal avalanche" refers to:
A sudden increase in temperature inside a neuron
A cascade of spikes that follows a power‑law size distribution
A type of synaptic plasticity
A computational algorithm for matrix inversion
Explanation - Neuronal avalanches are bursts of activity spanning multiple scales, indicative of critical dynamics in cortical networks.
Correct answer is: A cascade of spikes that follows a power‑law size distribution
Q.39 Which of the following is a common way to model the effect of myelination on action potential propagation speed?
Increasing membrane capacitance
Decreasing axial resistance in the internodal segments
Increasing leak conductance
Adding a delay term to the Hodgkin‑Huxley equations
Explanation - Myelination reduces axial resistance and capacitance in internodes, allowing faster saltatory conduction.
Correct answer is: Decreasing axial resistance in the internodal segments
Q.40 In a conductance‑based model, the total membrane current I_total is the sum of:
Leak current only
Capacitive current and all ionic currents
Only excitatory synaptic currents
Only inhibitory synaptic currents
Explanation - I_total = C_m dV/dt + Σ I_ion, where each ionic current includes its conductance and driving force.
Correct answer is: Capacitive current and all ionic currents
Q.41 Which of the following is a typical outcome when applying a low‑pass filter to a spike train?
Increased firing rate
Removal of high‑frequency components, smoothing the firing rate estimate
Conversion of spikes to binary code
Creation of new spikes
Explanation - Low‑pass filtering smooths the instantaneous firing rate by averaging over a time window, reducing rapid fluctuations.
Correct answer is: Removal of high‑frequency components, smoothing the firing rate estimate
Q.42 Which of the following best describes the role of NMDA receptors in synaptic plasticity?
They provide a fast, voltage‑independent current
They act as coincidence detectors due to voltage‑dependent Mg²⁺ block
They are purely inhibitory
They generate action potentials directly
Explanation - NMDA receptors require both glutamate binding and postsynaptic depolarization to relieve Mg²⁺ block, enabling calcium influx for plasticity.
Correct answer is: They act as coincidence detectors due to voltage‑dependent Mg²⁺ block
Q.43 In a spike‑based neural network model, the term "event‑driven" most closely refers to:
Updating all neuron states at every fixed time step
Updating neuron states only when a spike occurs
Using continuous differential equations for membrane potential
Simulating only one neuron at a time
Explanation - Event‑driven simulation processes changes only at spike times, reducing unnecessary calculations.
Correct answer is: Updating neuron states only when a spike occurs
Q.44 Which of the following is true about the relationship between the power spectrum of EEG and neuronal firing?
Higher power always indicates more spikes
Power in the gamma band (30‑80 Hz) is often linked to local synchronised firing
Alpha power reflects only muscle activity
Beta power is unrelated to cognition
Explanation - Gamma‑band oscillations are associated with local circuit synchrony and cognitive processes such as attention.
Correct answer is: Power in the gamma band (30‑80 Hz) is often linked to local synchronised firing
Q.45 Which of the following methods can be used to infer effective connectivity (directional influence) between brain regions from fMRI time series?
Granger causality analysis
Principal component analysis
Spike‑sorting
Wavelet denoising
Explanation - Granger causality tests whether past activity in one region improves prediction of another, implying directional influence.
Correct answer is: Granger causality analysis
Q.46 In a neural network trained with back‑propagation, the loss function gradient with respect to a weight w_ij is computed using:
Chain rule of calculus
Fourier transform
Laplace transform
Monte Carlo sampling
Explanation - Back‑propagation applies the chain rule to propagate error gradients from output back to each weight.
Correct answer is: Chain rule of calculus
Q.47 The term "synaptic delay" typically refers to:
The time for the presynaptic action potential to travel down the axon
The latency between presynaptic spike arrival and postsynaptic current onset
The refractory period of the postsynaptic neuron
The time needed for the neuron to recover after a burst
Explanation - Synaptic delay includes neurotransmitter release, diffusion, and receptor activation, usually 1‑2 ms in central synapses.
Correct answer is: The latency between presynaptic spike arrival and postsynaptic current onset
Q.48 Which of the following best describes a "spike‑train raster plot"?
A graph showing membrane voltage over time for a single neuron
A two‑dimensional plot where each row is a neuron and each dot marks a spike time
A histogram of inter‑spike intervals
A frequency spectrum of neural activity
Explanation - Raster plots visualize the timing of spikes across many neurons, facilitating detection of synchrony and patterns.
Correct answer is: A two‑dimensional plot where each row is a neuron and each dot marks a spike time
Q.49 Which of the following is a characteristic of a "winner‑take‑all" network?
All neurons fire at the same rate
Only the neuron (or group) with the strongest input remains active while others are suppressed
Neurons fire in a strict sequence
Synaptic weights are all equal
Explanation - Winner‑take‑all circuits implement competition, resulting in sparse, selective activity.
Correct answer is: Only the neuron (or group) with the strongest input remains active while others are suppressed
Q.50 In a conductance‑based neuron model, the term "shunting inhibition" refers to:
Hyperpolarizing current that moves V_m far below resting potential
Increase in membrane conductance that reduces the effect of excitatory inputs without large voltage change
A type of inhibition that only occurs during sleep
Inhibition that changes the reversal potential of excitatory synapses
Explanation - Shunting inhibition raises conductance, effectively dividing excitatory currents and dampening their impact, often near the resting potential.
Correct answer is: Increase in membrane conductance that reduces the effect of excitatory inputs without large voltage change
Q.51 Which of the following is a typical application of the Wilson‑Cowan equations?
Modeling individual action potentials
Describing the dynamics of interacting excitatory and inhibitory populations
Simulating detailed dendritic trees
Analyzing calcium imaging data directly
Explanation - Wilson‑Cowan equations capture the average firing rates of coupled excitatory and inhibitory neural populations.
Correct answer is: Describing the dynamics of interacting excitatory and inhibitory populations
Q.52 When using a Hidden Markov Model (HMM) to decode neural states, the hidden states represent:
Observed spike times
Unobserved brain states that generate observable spike patterns
Synaptic weights
Membrane potentials
Explanation - HMMs model a system where observable data (e.g., spikes) are generated by a sequence of hidden discrete states.
Correct answer is: Unobserved brain states that generate observable spike patterns
Q.53 The term "spike‑burst" in a neuron’s output is most closely associated with which of the following patterns?
A single isolated spike
A rapid series of spikes followed by a silent period
A constant firing rate of 5 Hz
No spiking activity
Explanation - Bursting consists of clusters of spikes (high frequency) separated by periods of quiescence.
Correct answer is: A rapid series of spikes followed by a silent period
Q.54 Which of the following best describes the purpose of a "learning rule" in a computational neural model?
To set the initial membrane potential
To specify how synaptic weights are updated based on activity
To determine the geometry of the neuron
To control the temperature of the simulation
Explanation - Learning rules (e.g., Hebbian, STDP) dictate the plasticity mechanisms that modify synaptic strengths.
Correct answer is: To specify how synaptic weights are updated based on activity
Q.55 In a spiking neural network, the "refractory period" ensures:
The neuron fires continuously without pause
A brief time after a spike when the neuron cannot fire again
Immediate re‑excitation after each spike
Synaptic weights are reset
Explanation - The refractory period imposes a minimum inter‑spike interval, reflecting biophysical limits of spike generation.
Correct answer is: A brief time after a spike when the neuron cannot fire again
Q.56 Which of the following is a typical consequence of increasing the strength of excitatory recurrent connections in a balanced network?
Network activity becomes completely silent
The network may transition to a high‑gain, possibly unstable regime
Inhibition is completely eliminated
Synaptic plasticity stops
Explanation - Stronger excitatory recurrence can push the network toward runaway excitation or oscillations if inhibition does not compensate.
Correct answer is: The network may transition to a high‑gain, possibly unstable regime
Q.57 In a computational model of the visual cortex, orientation selectivity is often generated by:
Randomly assigned synaptic weights
A Gabor‑like receptive field structure in the input layer
Uniform inhibition across all neurons
Absence of lateral connections
Explanation - Gabor filters emulate the elongated, oriented receptive fields of simple cells, leading to orientation selectivity.
Correct answer is: A Gabor‑like receptive field structure in the input layer
Q.58 Which of the following best describes the "curse of dimensionality" in neural data analysis?
Increasing dimensions always improves model accuracy
Data become sparse in high‑dimensional spaces, making statistical estimation difficult
Neurons cannot fire above a certain frequency
The brain cannot process high‑dimensional stimuli
Explanation - As dimensionality grows, the volume of space increases exponentially, requiring exponentially more data to achieve reliable estimates.
Correct answer is: Data become sparse in high‑dimensional spaces, making statistical estimation difficult
Q.59 When modeling a synapse with both AMPA and NMDA components, the total excitatory conductance g_E(t) is:
g_E = g_AMPA(t) + g_NMDA(t)
g_E = g_AMPA(t) × g_NMDA(t)
g_E = g_AMPA(t) - g_NMDA(t)
g_E = max(g_AMPA(t), g_NMDA(t))
Explanation - Excitatory synaptic conductance is the sum of fast AMPA and slower NMDA components.
Correct answer is: g_E = g_AMPA(t) + g_NMDA(t)
Q.60 In the context of neural coding, the "information rate" (bits/s) is maximized when:
The neuron fires at a constant rate
Spikes are perfectly predictable
Spikes are independent and highly variable
The firing pattern matches the stimulus entropy
Explanation - Maximum information transmission occurs when the output distribution mirrors the input (stimulus) entropy, optimizing coding efficiency.
Correct answer is: The firing pattern matches the stimulus entropy
Q.61 Which of the following is a common method to estimate the receptive field of a neuron from spike‑triggered averaging?
Compute the average stimulus preceding each spike
Calculate the Fourier transform of the spike train
Measure the membrane potential during silence
Count the number of spikes per minute
Explanation - Spike‑triggered averaging (STA) yields the linear component of the stimulus that most reliably precedes spikes.
Correct answer is: Compute the average stimulus preceding each spike
Q.62 Which of the following statements about the "Goldman‑Hodgkin‑Katz (GHK) voltage equation" is true?
It only applies to passive membranes
It calculates the reversal potential for a permeable ion based on concentration gradients and permeabilities
It predicts the exact timing of action potentials
It is identical to Ohm's law
Explanation - The GHK equation extends Nernst by weighting each ion's contribution by its relative permeability.
Correct answer is: It calculates the reversal potential for a permeable ion based on concentration gradients and permeabilities
Q.63 In a computational model, the term "noise floor" refers to:
The maximum possible firing rate of a neuron
The level of background activity below which signals cannot be reliably detected
The threshold voltage for spike initiation
The highest frequency component in the signal
Explanation - Noise floor defines the baseline noise level; signals must exceed it for detection.
Correct answer is: The level of background activity below which signals cannot be reliably detected
Q.64 Which of the following best characterizes a "feedforward" neural network architecture?
Connections only go from input to output without cycles
Neurons are connected in a loop
Signals travel backward from output to input
All neurons are mutually connected
Explanation - Feedforward networks have directed acyclic connections, preventing recurrent loops.
Correct answer is: Connections only go from input to output without cycles
Q.65 In the context of neural simulations, the term "time step" (Δt) primarily determines:
The number of neurons in the model
The numerical accuracy and stability of integration methods
The physical size of the brain region simulated
The amplitude of action potentials
Explanation - Δt sets the granularity of numerical integration; too large a step can cause instability or loss of precision.
Correct answer is: The numerical accuracy and stability of integration methods
Q.66 Which of the following is a hallmark of a “scale‑free” network often observed in brain connectivity graphs?
All nodes have the same degree
The degree distribution follows a power law
The network has a regular lattice structure
Every node connects to every other node
Explanation - Scale‑free networks have many nodes with few connections and few hubs with many connections, following a power‑law distribution.
Correct answer is: The degree distribution follows a power law
Q.67 When using the Euler method to integrate the differential equation dV/dt = f(V), the update rule for V after a time step Δt is:
V_{new} = V_{old} + f(V_{old}) Δt
V_{new} = V_{old} - f(V_{old}) Δt
V_{new} = V_{old} × f(V_{old}) Δt
V_{new} = f(V_{old}) / Δt
Explanation - Euler integration approximates the next value by adding the derivative times the time step to the current value.
Correct answer is: V_{new} = V_{old} + f(V_{old}) Δt
Q.68 In a model of short‑term facilitation, the synaptic efficacy typically:
Decreases with each successive spike
Increases transiently after presynaptic spikes and decays back
Remains constant regardless of spike history
Becomes zero after a single spike
Explanation - Facilitation temporarily raises release probability due to residual calcium, decaying with a characteristic time constant.
Correct answer is: Increases transiently after presynaptic spikes and decays back
Q.69 Which of the following is a typical reason to use a conductance‑based model rather than a current‑based model?
To reduce computational cost
To capture the voltage‑dependence of synaptic driving force
Because it does not require membrane capacitance
Because it eliminates the need for differential equations
Explanation - Conductance models reflect that synaptic current depends on the instantaneous membrane potential relative to reversal potentials.
Correct answer is: To capture the voltage‑dependence of synaptic driving force
Q.70 When fitting a logistic function to a neuron’s firing‑rate versus stimulus‑intensity curve, the parameter that determines the steepness of the transition is:
The offset (bias) term
The gain (slope) parameter
The maximum firing rate
The noise variance
Explanation - In a logistic function, the gain controls how rapidly the output changes with input around the threshold.
Correct answer is: The gain (slope) parameter
Q.71 Which of the following best defines the term "neural code"?
The DNA sequence that encodes ion channels
The set of rules by which neural activity represents information
The hardware architecture of a neuromorphic chip
The anatomical layout of brain regions
Explanation - Neural coding seeks to understand how patterns of spikes, rates, or synchrony encode sensory or motor information.
Correct answer is: The set of rules by which neural activity represents information
Q.72 In a computational model of a cortical column, lateral inhibition is typically implemented by:
Excitatory connections to distant neurons
Inhibitory connections from a neuron to its neighbors
Increasing the membrane capacitance
Removing all synaptic inputs
Explanation - Lateral inhibition suppresses neighboring activity via inhibitory interneurons, enhancing contrast.
Correct answer is: Inhibitory connections from a neuron to its neighbors
Q.73 Which of the following is a typical feature of a “spiking neural network” compared to a traditional artificial neural network?
Continuous activation functions
Discrete event‑driven communication via spikes
Back‑propagation through time
Fixed weight matrices
Explanation - Spiking networks communicate using time‑stamped spikes, capturing temporal dynamics absent in rate‑based ANNs.
Correct answer is: Discrete event‑driven communication via spikes
Q.74 Which of the following analysis techniques is most appropriate for identifying oscillatory components in a local field potential recording?
Wavelet transform
K‑means clustering
Linear regression
Decision tree classification
Explanation - Wavelet analysis provides time‑frequency decomposition, ideal for detecting transient oscillations.
Correct answer is: Wavelet transform
Q.75 In a model that includes both excitatory (E) and inhibitory (I) populations, the term "E/I balance" often refers to:
The ratio of the number of excitatory to inhibitory neurons
The equality of total excitatory and inhibitory conductances at any moment
The difference in membrane capacitance between E and I cells
The absolute silence of both populations
Explanation - E/I balance is achieved when the net excitatory and inhibitory inputs cancel on average, stabilizing network activity.
Correct answer is: The equality of total excitatory and inhibitory conductances at any moment
Q.76 Which of the following is true about the “softmax” function commonly used in neural network classifiers?
It outputs binary values (0 or 1)
It converts a vector of real numbers into a probability distribution
It reduces the dimensionality of the input
It adds Gaussian noise to the inputs
Explanation - Softmax exponentiates each input and normalizes by the sum, yielding values between 0 and 1 that sum to 1.
Correct answer is: It converts a vector of real numbers into a probability distribution
Q.77 When modeling a neuron with a refractory period τ_ref, the simplest way to implement it in a discrete‑time simulation is to:
Set the membrane potential to zero for τ_ref steps after each spike
Ignore any incoming spikes during τ_ref steps
Add a constant current during τ_ref
Increase the synaptic weight during τ_ref
Explanation - A refractory period prevents the neuron from spiking again; in simulations this is often done by disabling spike generation for a set number of steps.
Correct answer is: Ignore any incoming spikes during τ_ref steps
Q.78 Which of the following best describes the purpose of "cross‑validation" in the context of building a neural decoding model?
To increase the size of the dataset
To assess how well the model generalizes to unseen data
To compute the exact solution of the model equations
To visualize spike rasters
Explanation - Cross‑validation partitions data into training and testing sets, providing an unbiased estimate of predictive performance.
Correct answer is: To assess how well the model generalizes to unseen data
Q.79 In a computational model of a neural population, a "limit cycle" refers to:
A stable equilibrium where activity ceases
A periodic trajectory that the system repeatedly follows
A chaotic attractor with no repeating pattern
A point where all neurons fire synchronously
Explanation - A limit cycle is a closed orbit in phase space, representing sustained oscillations.
Correct answer is: A periodic trajectory that the system repeatedly follows
Q.80 Which of the following statements about the relationship between fMRI BOLD signal and neural activity is most accurate?
BOLD directly measures action potentials
BOLD reflects local field potentials and synaptic activity more than spiking
BOLD is unrelated to neuronal metabolism
BOLD signals have millisecond temporal resolution
Explanation - The BOLD response is driven by changes in blood flow and oxygenation linked to synaptic and dendritic activity rather than direct spiking.
Correct answer is: BOLD reflects local field potentials and synaptic activity more than spiking
Q.81 When applying the Hodgkin‑Huxley formalism to a myelinated axon, which modification is typically necessary?
Add a large leak conductance throughout the axon
Introduce discrete nodes of Ranvier with active ion channels and passive internodes
Replace all sodium channels with potassium channels
Set the membrane capacitance to zero
Explanation - Myelinated axons are modeled as active nodes separated by passive, low‑capacitance internodes to capture saltatory conduction.
Correct answer is: Introduce discrete nodes of Ranvier with active ion channels and passive internodes
Q.82 In a recurrent network, the term "Hebbian learning with weight normalization" is used to:
Prevent weights from growing without bound by scaling them after each update
Make all weights equal to zero
Remove the Hebbian term from the learning rule
Increase the learning rate indefinitely
Explanation - Weight normalization rescales synaptic strengths to maintain overall stability while preserving relative differences.
Correct answer is: Prevent weights from growing without bound by scaling them after each update
Q.83 Which of the following is the most appropriate way to model stochastic ion channel noise in a single‑compartment neuron?
Add a deterministic current term
Introduce a Gaussian white noise term into the membrane equation
Increase the membrane capacitance
Remove the leak conductance
Explanation - Channel noise can be approximated by adding a zero‑mean Gaussian noise current, reflecting random opening/closing of channels.
Correct answer is: Introduce a Gaussian white noise term into the membrane equation
Q.84 In the context of neural data, the "receiver operating characteristic (ROC) curve" is used to:
Measure the speed of spike propagation
Evaluate the trade‑off between true‑positive and false‑positive rates of a classifier
Compute the power spectrum of a signal
Determine the refractory period
Explanation - ROC curves plot sensitivity versus (1‑specificity), summarizing classifier performance across thresholds.
Correct answer is: Evaluate the trade‑off between true‑positive and false‑positive rates of a classifier
Q.85 Which of the following best describes the "integrate‑and‑fire" mechanism of a neuron?
The neuron integrates incoming currents until a threshold is reached, then fires a spike and resets
The neuron fires continuously regardless of input
The neuron integrates voltage without any threshold
The neuron only responds to inhibitory inputs
Explanation - The integrate‑and‑fire model accumulates input charge, emits a spike when threshold is crossed, and then resets.
Correct answer is: The neuron integrates incoming currents until a threshold is reached, then fires a spike and resets
Q.86 In computational neuroscience, a "phase response curve" (PRC) characterizes:
How a neuron's firing frequency changes with temperature
The change in spike timing caused by a brief perturbation at different phases of the cycle
The distribution of synaptic weights
The voltage dependence of ion channels
Explanation - PRCs map the phase shift induced by a stimulus applied at various points in the inter‑spike interval.
Correct answer is: The change in spike timing caused by a brief perturbation at different phases of the cycle
Q.87 When analyzing spike train synchrony, the "joint peri‑stimulus time histogram" (JPSTH) is used to:
Estimate the average firing rate of a single neuron
Visualize the time‑locked correlation between two simultaneously recorded neurons
Compute the power spectrum of a local field potential
Determine the membrane capacitance
Explanation - JPSTH displays the joint firing probability as a function of the time lag between two neurons, revealing synchrony patterns.
Correct answer is: Visualize the time‑locked correlation between two simultaneously recorded neurons
Q.88 In a conductance‑based model, the reversal potential for chloride (Cl⁻) is typically around:
-70 mV
-50 mV
0 mV
+60 mV
Explanation - Cl⁻ reversal potential is close to the resting membrane potential, making GABAergic inhibition typically hyperpolarizing.
Correct answer is: -70 mV
Q.89 Which of the following statements about the “canonical microcircuit” in neocortex is correct?
It consists only of excitatory pyramidal cells
It includes a specific pattern of feedforward excitation and feedback inhibition across layers
It lacks any inhibitory interneurons
It is only present in the hippocampus
Explanation - The canonical microcircuit describes a stereotyped arrangement of excitatory and inhibitory connections across cortical layers that supports information processing.
Correct answer is: It includes a specific pattern of feedforward excitation and feedback inhibition across layers
Q.90 In a model that incorporates spike‑frequency adaptation, the adaptation current typically:
Increases with each spike and decays slowly
Decreases with each spike
Remains constant over time
Is independent of spiking activity
Explanation - Adaptation currents (e.g., K⁺-mediated) build up with spiking, causing gradual firing rate reduction, and decay over longer time constants.
Correct answer is: Increases with each spike and decays slowly
Q.91 Which of the following best describes the role of dendritic spikes in neuronal computation?
They have no effect on the soma
They can locally amplify synaptic inputs and influence somatic output nonlinearly
They only occur in inhibitory neurons
They are identical to axonal action potentials
Explanation - Dendritic spikes provide a mechanism for nonlinear integration of inputs, affecting the overall firing probability of the neuron.
Correct answer is: They can locally amplify synaptic inputs and influence somatic output nonlinearly
Q.92 When fitting a model to neural data, the "Akaike Information Criterion (AIC)" is used to:
Maximize the likelihood regardless of model complexity
Balance model fit quality against the number of parameters
Compute the correlation coefficient
Determine the spike threshold
Explanation - AIC penalizes excessive parameters to avoid overfitting, selecting models that achieve a good trade‑off between accuracy and simplicity.
Correct answer is: Balance model fit quality against the number of parameters
Q.93 Which of the following is a common way to model the effect of neuromodulators (e.g., dopamine) on synaptic plasticity in computational models?
By adding a constant current to all neurons
By scaling the learning rate or plasticity windows based on modulatory signals
By removing all inhibitory connections
By increasing the membrane capacitance globally
Explanation - Neuromodulators often modulate the magnitude or timing of plasticity rules, effectively altering learning rates in models.
Correct answer is: By scaling the learning rate or plasticity windows based on modulatory signals
Q.94 In a spiking neural network simulation, the "synaptic delay" is most accurately implemented by:
Adding a fixed number of simulation time steps before a spike influences the postsynaptic neuron
Increasing the membrane resistance
Changing the reversal potential of the synapse
Altering the spike threshold
Explanation - Synaptic delay is modeled by queuing the spike for a set number of time steps before it triggers the postsynaptic conductance change.
Correct answer is: Adding a fixed number of simulation time steps before a spike influences the postsynaptic neuron
Q.95 Which of the following best describes a "random walk" model for neural firing rates?
Deterministic increase to a fixed point
Stochastic accumulation of noise leading to threshold crossing
Periodic oscillation with constant amplitude
Immediate reset after each spike
Explanation - Random walk models treat membrane potential as a noisy process that drifts until it reaches a firing threshold.
Correct answer is: Stochastic accumulation of noise leading to threshold crossing
Q.96 In a model of a cortical microcircuit, feedforward inhibition typically results in:
Delayed excitation
A rapid suppression of excitatory input, sharpening temporal response
Elimination of all spikes
Increase in synaptic weight
Explanation - Feedforward inhibition provides fast inhibitory input that curtails excitatory responses, improving temporal precision.
Correct answer is: A rapid suppression of excitatory input, sharpening temporal response
Q.97 Which of the following is a standard measure of synchrony that accounts for the phase of oscillations across trials?
Mean firing rate
Phase‑locking value (PLV)
Spike count histogram
Inter‑spike interval variance
Explanation - PLV quantifies the consistency of spike or signal phase relative to an oscillatory reference across trials.
Correct answer is: Phase‑locking value (PLV)
Q.98 In a conductance‑based model with voltage‑dependent NMDA receptors, the current through NMDA channels depends on:
Only the presynaptic firing rate
Both the membrane voltage (due to Mg²⁺ block) and the neurotransmitter concentration
Only the temperature of the tissue
Only the leak conductance
Explanation - NMDA receptors exhibit voltage‑dependent relief of Mg²⁺ block and require glutamate binding, making their current voltage‑ and ligand‑dependent.
Correct answer is: Both the membrane voltage (due to Mg²⁺ block) and the neurotransmitter concentration
Q.99 When using the Levenberg‑Marquardt algorithm for fitting a neuronal model, the algorithm combines features of:
Gradient descent and Newton's method
Simulated annealing and genetic algorithms
Monte Carlo sampling and Bayesian inference
Fourier analysis and wavelet transforms
Explanation - Levenberg‑Marquardt interpolates between gradient descent (for stability) and Newton's method (for speed) to solve nonlinear least‑squares problems.
Correct answer is: Gradient descent and Newton's method
Q.100 In a model of a sensory neuron, the term "tuning curve" refers to:
The time course of a spike after a stimulus
The relationship between stimulus intensity (or feature) and firing rate
The anatomical shape of the neuron
The synaptic weight distribution
Explanation - Tuning curves characterize how a neuron's firing rate varies with a specific stimulus parameter (e.g., orientation, frequency).
Correct answer is: The relationship between stimulus intensity (or feature) and firing rate
Q.101 Which of the following is a key assumption behind the use of linear‑filter models (e.g., STA) for neural encoding?
Neurons respond nonlinearly to all inputs
The relationship between stimulus and response is linear and time‑invariant
Synaptic conductances are zero
Noise is deterministic
Explanation - Linear‑filter models assume that the neuron's response can be described by a convolution of the stimulus with a linear kernel.
Correct answer is: The relationship between stimulus and response is linear and time‑invariant
Q.102 In computational models of learning, the term "eligibility trace" is used to:
Record the recent history of presynaptic activity for credit assignment
Measure the membrane capacitance
Set the firing threshold of a neuron
Determine the axonal length
Explanation - Eligibility traces store temporally decaying information about activity, allowing delayed reinforcement signals to modify synapses.
Correct answer is: Record the recent history of presynaptic activity for credit assignment
Q.103 Which of the following is a typical characteristic of a "small‑world" network observed in brain connectivity?
High clustering coefficient and short average path length
All nodes have identical degree
Only nearest‑neighbor connections
Fully random connections without clustering
Explanation - Small‑world networks combine local clustering with global efficiency, a pattern found in many neural systems.
Correct answer is: High clustering coefficient and short average path length
Q.104 When modeling a neuron with stochastic spike generation (Poisson process), the probability of observing k spikes in a time window Δt with rate λ is given by:
P(k) = (λΔt)^k e^{-λΔt} / k!
P(k) = λ / (Δt + k)
P(k) = k! / (λΔt)^k
P(k) = e^{-λΔt} / (λΔt)^k
Explanation - The Poisson distribution gives the probability of k events occurring in a fixed interval with mean λΔt.
Correct answer is: P(k) = (λΔt)^k e^{-λΔt} / k!
Q.105 Which of the following is true about the relationship between membrane resistance (R_m) and input resistance measured at the soma?
They are always equal
Input resistance is typically larger due to dendritic filtering
Input resistance is always smaller because of capacitive effects
Input resistance does not depend on R_m
Explanation - Dendritic trees add series resistance, making the measured somatic input resistance larger than the local membrane resistance.
Correct answer is: Input resistance is typically larger due to dendritic filtering
Q.106 In a neural network model, the term "weight decay" refers to:
Increasing the learning rate over time
A regularization technique that gradually reduces synaptic strengths
Removing all inhibitory connections
Setting all weights to zero after each epoch
Explanation - Weight decay adds a penalty proportional to the square of the weight, encouraging smaller weights and preventing overfitting.
Correct answer is: A regularization technique that gradually reduces synaptic strengths
Q.107 Which of the following best describes the purpose of a "spike‑generator" in a neuromorphic hardware platform?
To provide deterministic voltage waveforms
To emit timed electrical pulses that emulate biological spikes for testing circuits
To increase the capacitance of the chip
To store synaptic weights
Explanation - Spike‑generators produce programmable spike trains that can drive neuromorphic circuits, mimicking neuronal output.
Correct answer is: To emit timed electrical pulses that emulate biological spikes for testing circuits
Q.108 When analyzing multi‑unit activity, the term "multi‑unit firing rate" typically refers to:
The average firing rate of a single neuron
The combined firing rate of all recorded units within a small time window
The membrane potential of the neuron
The number of synapses per neuron
Explanation - Multi‑unit activity aggregates spikes from multiple nearby neurons, providing a population-level firing estimate.
Correct answer is: The combined firing rate of all recorded units within a small time window
Q.109 Which of the following is a typical signature of a "bursting" regime in a neuronal phase‑plane plot?
A single fixed point
A closed trajectory (limit cycle) with fast spikes followed by a slow manifold
A straight line through the origin
No trajectory at all
Explanation - Bursting dynamics involve fast spiking on a limit cycle interleaved with slow recovery dynamics, visible as a two‑timescale trajectory.
Correct answer is: A closed trajectory (limit cycle) with fast spikes followed by a slow manifold
Q.110 In a model of a neuron receiving both excitatory and inhibitory Poisson inputs, the total variance of the membrane potential scales with:
The sum of the rates of excitatory and inhibitory inputs
Only the excitatory input rate
Only the inhibitory input rate
The product of the two rates
Explanation - Each Poisson input contributes independently to membrane noise; variances add, so total variance is proportional to the sum of rates.
Correct answer is: The sum of the rates of excitatory and inhibitory inputs
Q.111 Which of the following best explains why the Bienenstock‑Cooper‑Munro (BCM) learning rule includes a sliding threshold?
To keep synaptic weights constant over time
To allow the neuron to adapt its plasticity based on recent activity levels
To make the rule independent of input statistics
To enforce binary synapses
Explanation - The sliding threshold ensures that the neuron undergoes potentiation or depression depending on whether its recent activity is above or below a moving average.
Correct answer is: To allow the neuron to adapt its plasticity based on recent activity levels
Q.112 In computational models, the term "stochastic resonance" refers to:
Noise destroying the signal completely
The phenomenon where a certain level of noise enhances the detection of weak signals
A deterministic oscillation at a fixed frequency
A type of synaptic plasticity
Explanation - Stochastic resonance occurs when adding optimal noise improves the signal‑to‑noise ratio for subthreshold stimuli.
Correct answer is: The phenomenon where a certain level of noise enhances the detection of weak signals