L4
Course Structure
- Semester Overview
- Structured week by week, covering different topics in neural data science.
- Each session includes lectures, exercises, and discussions to reinforce learning.
Lecture Overview for Lecture 4
- Recap of Lecture 3
- Beyond Rate Coding
- Sampling from Probability Distributions
- Gamma Process
- Correlations
- Population Analyses
- The Problem of Averaging
- Z-Scores
Key Concepts from Lecture 4
Fano Factor
- Definition: Measures variability in spike count across trials.
- Formula: Fano Factor = (Variance of Spike Count) / (Mean Spike Count)
- Application: Assess variability in neural responses over shorter time windows.
- Example: Used in Churchland et al. (2010) to demonstrate how stimulus onset can reduce variability in neural activity.
Inter-Spike Interval (ISI)
- Distributions: Analysis of ISI is crucial for understanding neural firing patterns.
- Utility: Helps interpret empirical spike trains and identify if they fit a given distribution (e.g., gamma).
Coefficient of Variation (CV)
- Definition: Measures variability in ISI.
- Formula: CV = (Standard Deviation of ISI) / (Mean ISI)
Poisson Process
- Description: A model for describing the number of events (spikes) occurring in a fixed period.
- Formula: PT[n] = (rT)^n * e^(-rT) / n!
- Characteristics: Suitable for modeling random spike generation in neurons having a constant average rate.
Sampling from Probability Distributions
- Probability Density Functions (PDFs): Describe continuous variables' likelihoods. Integral of PDF = 1.
- Normal Distribution: Key example with mean (μ) and variance (σ²).
- Cumulative Distribution Function (CDF): Represents the probability that a random variable takes on a value less than or equal to x.
- Inversion Method: A sampling technique applied to generate samples from a given distribution.
Gamma Distribution
- Characteristics: A complex family of distributions with various shapes; the exponential distribution is a special case.
- Applications: Generates more regular or variable spike trains than Poisson spike trains.
Correlation Measures
- Autocorrelation: Examines probability of neurons firing after a certain interval. Estimates periodic firing and activity patterns.
- Cross-Correlation: Evaluates firing patterns between different neurons, revealing resource-sharing or coordinated firing behaviors.
Population Analyses
The Problem of Averaging
- Heterogeneity in Neurons: Different neurons respond variably due to factors like input strength and firing rate.
- Z-Scores for Normalization: Allow for fair comparison across different neuron types by standardizing firing rates:
- Formula: z = (x - μ) / σ
Challenges with Averaging
- Insights from Raster Plots: Reveal variability masked by simple averaging processes.
- Loss of Information: Simple averaging of firing rates may obscure meaningful differences in neuron activity.
Revisiting Concepts for Understanding Neural Coding
- Spike Train Analysis: Techniques include examining ISI distributions, applying correlation statistics, and processing complex firing patterns across trials.
Summary of Learning Outcomes
- Explain and apply analysis for spike train variability and correlations.
- Describe and generate artificial spike trains.
- Utilize auto- and cross-correlations effectively.
- Analyze activity across multiple neurons using advanced methods.
Questions for Revision
- Sketch a raster plot reflecting changes due to a stimulus.
- Discuss how burst firing changes ISI distributions.
- Evaluate the effects of refractory periods on ISI histograms in relation to Poisson processes.
- Define the implications of a z-score of -1, 0, and +1 in firing rates.