Information Theory Module

Information theory functions for DRIADA.

This module provides various information-theoretic measures including mutual information, entropy, and conditional information measures.

Information-theoretic measures for analyzing neural data, including entropy estimation, mutual information computation, and advanced corrections for finite data.

Module Components

• Core Information Theory Classes
• Entropy Estimation
• Mutual Information Functions
• Information Estimators
• Information Theory Utilities

Quick Links

Core Data Structures

• TimeSeries - Single time series container

• MultiTimeSeries - Multi-dimensional time series

• Core Information Theory Classes - Base classes and main MI functions

Mutual Information

• get_mi() - Main MI computation function

• conditional_mi() - Conditional MI

• interaction_information() - Multi-variable interactions

• Mutual Information Functions - All MI-related functions

Entropy Estimation

• Entropy Estimation - Discrete and continuous entropy (JIT-optimized)

Estimators

• Information Estimators - GCMI (Gaussian copula) and KSG (k-nearest neighbors)

Usage Example

from driada.information import TimeSeries, get_mi
import numpy as np

# Create time series with matching lengths
n_samples = 1000

# Generate example data
spike_counts = np.random.poisson(3, n_samples)  # Discrete spike counts
position = np.cumsum(np.random.randn(n_samples) * 0.1)  # Continuous position

# Ensure data has no extreme outliers for GCMI
position = np.clip(position, np.percentile(position, 1), np.percentile(position, 99))

neural_data = TimeSeries(spike_counts, discrete=True)
behavior = TimeSeries(position, discrete=False)

# Compute mutual information
mi_value = get_mi(neural_data, behavior, estimator='gcmi')

# Conditional MI (X must be continuous)
from driada.information import conditional_mi
speed = np.abs(np.diff(position, prepend=position[0]))  # Speed from position
speed_ts = TimeSeries(speed[:n_samples], discrete=False)
cmi = conditional_mi(behavior, neural_data, speed_ts)  # I(position; spikes | speed)