Crate xlog_neural 

Neural network integration for XLOG probabilistic logic programs.

This crate provides the infrastructure for integrating PyTorch neural networks with XLOG’s probabilistic inference engine, following the DeepProbLog paradigm.

Architecture

The neural integration consists of:

• NetworkRegistry: Central registry managing all neural networks
• NetworkHandle: Holds PyTorch module, optimizer, and configuration
• NetworkConfig: Configuration options for network behavior

Features

• python - Enable Python interop via PyO3. Required for actual PyTorch integration.

Example

use xlog_neural::{NetworkRegistry, NetworkConfig};

let mut registry = NetworkRegistry::new();
registry.register(NetworkConfig::default("mnist_net"));

// Set all networks to training mode
registry.set_train_mode(true);

Re-exports

pub use batch::BatchCollector;

pub use batch::BatchMapping;

pub use batch::BatchResult;

pub use batch::NeuralCall;

pub use bridge::ADProbability;

pub use bridge::CircuitLeaf;

pub use bridge::NeuralBridge;

pub use bridge::NeuralOutput;

pub use handle::EmbeddingHandle;

pub use handle::NetworkHandle;

pub use registry::NetworkConfig;

pub use registry::NetworkRegistry;

pub use tensor_source::TensorMetadata;

pub use tensor_source::TensorSourceError;

pub use tensor_source::TensorSourceRegistry;

pub use pyo3;

Modules

batch

Batched Neural Evaluation

bridge

Neural → Probability Bridge

handle

Network handle for managing PyTorch modules.

registry

Network registry for managing registered neural networks.

tensor_source

Tensor Source Registry

Enums

NeuralError

Error types for neural network operations

Type Aliases

NeuralResult

Result type for neural operations