Struct DeepBoltzmannMachine 

pub struct DeepBoltzmannMachine { /* private fields */ }

Deep Boltzmann Machine with multiple layers

Implementations

impl DeepBoltzmannMachine

pub fn new( layer_sizes: Vec<usize>, temperature: f64, learning_rate: f64, ) -> Result<Self>

Create a new Deep Boltzmann Machine

Examples found in repository

examples/quantum_boltzmann.rs (lines 131-135)

fn deep_boltzmann_demo() -> Result<()> {
    // Create a 3-layer DBM
    let layer_sizes = vec![8, 4, 2];
    let mut dbm = DeepBoltzmannMachine::new(
        layer_sizes.clone(),
        1.0,  // temperature
        0.01, // learning rate
    )?;

    println!("   Created Deep Boltzmann Machine:");
    println!("   - Architecture: {layer_sizes:?}");
    println!("   - Total layers: {}", dbm.rbms().len());

    // Generate hierarchical data
    let data = generate_hierarchical_data(300, 8);

    // Layer-wise pretraining
    println!("\n   Performing layer-wise pretraining...");
    dbm.pretrain(
        &data, 50, // epochs per layer
        30, // batch size
    )?;

    println!("\n   Pretraining complete!");
    println!("   Each layer learned increasingly abstract features");

    Ok(())
}

pub fn pretrain( &mut self, data: &Array2<f64>, epochs_per_layer: usize, batch_size: usize, ) -> Result<()>

Layer-wise pretraining

Examples found in repository

examples/quantum_boltzmann.rs (lines 146-149)

fn deep_boltzmann_demo() -> Result<()> {
    // Create a 3-layer DBM
    let layer_sizes = vec![8, 4, 2];
    let mut dbm = DeepBoltzmannMachine::new(
        layer_sizes.clone(),
        1.0,  // temperature
        0.01, // learning rate
    )?;

    println!("   Created Deep Boltzmann Machine:");
    println!("   - Architecture: {layer_sizes:?}");
    println!("   - Total layers: {}", dbm.rbms().len());

    // Generate hierarchical data
    let data = generate_hierarchical_data(300, 8);

    // Layer-wise pretraining
    println!("\n   Performing layer-wise pretraining...");
    dbm.pretrain(
        &data, 50, // epochs per layer
        30, // batch size
    )?;

    println!("\n   Pretraining complete!");
    println!("   Each layer learned increasingly abstract features");

    Ok(())
}

pub fn rbms(&self) -> &[QuantumRBM]

Get the RBMs

Examples found in repository

examples/quantum_boltzmann.rs (line 139)

fn deep_boltzmann_demo() -> Result<()> {
    // Create a 3-layer DBM
    let layer_sizes = vec![8, 4, 2];
    let mut dbm = DeepBoltzmannMachine::new(
        layer_sizes.clone(),
        1.0,  // temperature
        0.01, // learning rate
    )?;

    println!("   Created Deep Boltzmann Machine:");
    println!("   - Architecture: {layer_sizes:?}");
    println!("   - Total layers: {}", dbm.rbms().len());

    // Generate hierarchical data
    let data = generate_hierarchical_data(300, 8);

    // Layer-wise pretraining
    println!("\n   Performing layer-wise pretraining...");
    dbm.pretrain(
        &data, 50, // epochs per layer
        30, // batch size
    )?;

    println!("\n   Pretraining complete!");
    println!("   Each layer learned increasingly abstract features");

    Ok(())
}