Skip to content

Bayesian Network I/O

The I/O module provides functionality for reading and writing Bayesian Networks in various standard formats. This enables interoperability with other tools and persistence of network models.

Overview

The causaliq_core.bn.io module supports multiple file formats commonly used for Bayesian Networks:

  • DSC Format: CausalIQ's native format for network specification
  • XDSL Format: GeNIe/SMILE's XML-based format for complete networks
  • Common Interface: Unified functions for format-agnostic I/O operations

Supported Formats

DSC Format

A text-based format for specifying Bayesian Network structures and parameters.

Features:

  • Human-readable network specification
  • Support for both discrete and continuous distributions
  • Compact representation of network structure
  • Easy manual editing and version control

XDSL Format

An XML-based format used by GeNIe and SMILE software.

Features:

  • Complete network serialization including visual layout
  • Industry-standard format for tool interoperability
  • Support for complex network metadata
  • Preservation of all network properties

Main Functions

The module provides convenient top-level functions:

from causaliq_core.bn.io import read_bn, write_bn

# Read BN from file (format auto-detected)
bn = read_bn('network.dsc')     # DSC format
bn = read_bn('network.xdsl')    # XDSL format

# Write BN to file (format determined by extension)  
write_bn(bn, 'output.dsc')      # Save as DSC
write_bn(bn, 'output.xdsl')     # Save as XDSL

Format Detection

The I/O system automatically detects formats based on:

  • File Extension: .dsc for DSC format, .xdsl for XDSL format
  • Content Analysis: Inspection of file headers and structure
  • Explicit Specification: Manual format specification when needed

Usage Patterns

Loading Networks

from causaliq_core.bn.io import read_bn

# Load from different formats
dsc_network = read_bn('models/weather.dsc')
xdsl_network = read_bn('models/medical.xdsl') 

# Verify loaded network
print(f"Nodes: {list(dsc_network.dag.nodes)}")
print(f"Edges: {list(dsc_network.dag.edges)}")

Saving Networks

from causaliq_core.bn.io import write_bn
from causaliq_core.bn import BN

# Create or modify network
# ... bn creation code ...

# Save in different formats
write_bn(bn, 'output/final_model.dsc')     # DSC format
write_bn(bn, 'output/final_model.xdsl')    # XDSL format

# Backup with timestamp
import datetime
timestamp = datetime.datetime.now().strftime('%Y%m%d_%H%M%S')
write_bn(bn, f'backups/model_{timestamp}.dsc')

Format Conversion

from causaliq_core.bn.io import read_bn, write_bn

# Convert between formats
bn = read_bn('input.dsc')       # Read DSC
write_bn(bn, 'output.xdsl')     # Write XDSL

# Batch conversion
import glob
for dsc_file in glob.glob('*.dsc'):
    bn = read_bn(dsc_file)
    xdsl_file = dsc_file.replace('.dsc', '.xdsl')
    write_bn(bn, xdsl_file)

Error Handling

The I/O system provides robust error handling:

  • Format Validation: Checks for valid file formats and structure
  • Parse Errors: Clear messages for syntax and content errors
  • Missing Files: Helpful error messages for file access issues
  • Recovery: Graceful handling of partial or corrupted files

Performance Considerations

  • Lazy Loading: Large networks are loaded efficiently
  • Memory Management: Minimal memory footprint during I/O operations
  • Streaming: Support for large files that don't fit in memory
  • Caching: Intelligent caching of frequently accessed networks

API Reference

io

I/O module for Bayesian Network file formats.

Modules:

Functions:

  • read_bn –

    Read a Bayesian Network from a file, automatically detecting format.

  • write_bn –

    Write BN to a DSC or XDSL format file.

read_bn 

read_bn(path: str, correct: bool = False) -> BN

Read a Bayesian Network from a file, automatically detecting format.

Supports: - .dsc files (DSC format) - .xdsl files (XDSL format)

Parameters:

  • path
    (str) –

    Path to DSC/XDSL file.

  • correct
    (bool, default: False ) –

    Whether to correct probabilities that do not sum to 1 (XDSL files only).

Returns:

  • BN –

    Bayesian Network specified in file.

Raises:

  • TypeError –

    If path is not a string.

  • ValueError –

    If path suffix is not "dsc" or "xdsl".

  • FileNotFoundError –

    If file does not exist.

  • FileFormatError –

    If file contents not valid.

write_bn 

write_bn(bn: BN, path: str) -> None

Write BN to a DSC or XDSL format file.

Parameters:

  • bn
    (BN) –

    Bayesian Network to write.

  • path
    (str) –

    Path to file.

Raises:

  • ValueError –

    If suffix not ".dsc" or ".xdsl".

  • FileNotFoundError –

    If file location nonexistent.