Struct QuantumScoreDiffusion 

pub struct QuantumScoreDiffusion { /* private fields */ }

Quantum Score-Based Diffusion

Implementations

impl QuantumScoreDiffusion

pub fn new( data_dim: usize, num_qubits: usize, num_noise_levels: usize, ) -> Result<Self>

Create new score-based diffusion model

Examples found in repository

examples/quantum_diffusion.rs (lines 178-182)

fn score_diffusion_demo() -> Result<()> {
    // Create score-based model
    let model = QuantumScoreDiffusion::new(
        2,  // data dimension
        4,  // num qubits
        10, // noise levels
    )?;

    println!("   Created quantum score-based diffusion model");
    println!("   - Noise levels: {:?}", model.noise_levels());

    // Test score estimation
    let x = Array1::from_vec(vec![0.5, -0.3]);
    let noise_level = 0.1;

    let score = model.estimate_score(&x, noise_level)?;
    println!("\n   Score estimation:");
    println!("   - Input: [{:.3}, {:.3}]", x[0], x[1]);
    println!("   - Noise level: {noise_level:.3}");
    println!("   - Estimated score: [{:.3}, {:.3}]", score[0], score[1]);

    // Langevin sampling
    println!("\n   Langevin sampling:");
    let init = Array1::from_vec(vec![2.0, 2.0]);
    let num_steps = 100;
    let step_size = 0.01;

    let sample = model.langevin_sample(init.clone(), noise_level, num_steps, step_size)?;

    println!("   - Initial: [{:.3}, {:.3}]", init[0], init[1]);
    println!(
        "   - After {} steps: [{:.3}, {:.3}]",
        num_steps, sample[0], sample[1]
    );
    println!(
        "   - Distance moved: {:.3}",
        (sample[0] - init[0]).hypot(sample[1] - init[1])
    );

    Ok(())
}

pub fn estimate_score( &self, x: &Array1<f64>, noise_level: f64, ) -> Result<Array1<f64>>

Estimate score (gradient of log density)

Examples found in repository

examples/quantum_diffusion.rs (line 191)

fn score_diffusion_demo() -> Result<()> {
    // Create score-based model
    let model = QuantumScoreDiffusion::new(
        2,  // data dimension
        4,  // num qubits
        10, // noise levels
    )?;

    println!("   Created quantum score-based diffusion model");
    println!("   - Noise levels: {:?}", model.noise_levels());

    // Test score estimation
    let x = Array1::from_vec(vec![0.5, -0.3]);
    let noise_level = 0.1;

    let score = model.estimate_score(&x, noise_level)?;
    println!("\n   Score estimation:");
    println!("   - Input: [{:.3}, {:.3}]", x[0], x[1]);
    println!("   - Noise level: {noise_level:.3}");
    println!("   - Estimated score: [{:.3}, {:.3}]", score[0], score[1]);

    // Langevin sampling
    println!("\n   Langevin sampling:");
    let init = Array1::from_vec(vec![2.0, 2.0]);
    let num_steps = 100;
    let step_size = 0.01;

    let sample = model.langevin_sample(init.clone(), noise_level, num_steps, step_size)?;

    println!("   - Initial: [{:.3}, {:.3}]", init[0], init[1]);
    println!(
        "   - After {} steps: [{:.3}, {:.3}]",
        num_steps, sample[0], sample[1]
    );
    println!(
        "   - Distance moved: {:.3}",
        (sample[0] - init[0]).hypot(sample[1] - init[1])
    );

    Ok(())
}

pub fn langevin_sample( &self, init: Array1<f64>, noise_level: f64, num_steps: usize, step_size: f64, ) -> Result<Array1<f64>>

Langevin dynamics sampling

Examples found in repository

examples/quantum_diffusion.rs (line 203)

fn score_diffusion_demo() -> Result<()> {
    // Create score-based model
    let model = QuantumScoreDiffusion::new(
        2,  // data dimension
        4,  // num qubits
        10, // noise levels
    )?;

    println!("   Created quantum score-based diffusion model");
    println!("   - Noise levels: {:?}", model.noise_levels());

    // Test score estimation
    let x = Array1::from_vec(vec![0.5, -0.3]);
    let noise_level = 0.1;

    let score = model.estimate_score(&x, noise_level)?;
    println!("\n   Score estimation:");
    println!("   - Input: [{:.3}, {:.3}]", x[0], x[1]);
    println!("   - Noise level: {noise_level:.3}");
    println!("   - Estimated score: [{:.3}, {:.3}]", score[0], score[1]);

    // Langevin sampling
    println!("\n   Langevin sampling:");
    let init = Array1::from_vec(vec![2.0, 2.0]);
    let num_steps = 100;
    let step_size = 0.01;

    let sample = model.langevin_sample(init.clone(), noise_level, num_steps, step_size)?;

    println!("   - Initial: [{:.3}, {:.3}]", init[0], init[1]);
    println!(
        "   - After {} steps: [{:.3}, {:.3}]",
        num_steps, sample[0], sample[1]
    );
    println!(
        "   - Distance moved: {:.3}",
        (sample[0] - init[0]).hypot(sample[1] - init[1])
    );

    Ok(())
}

pub fn noise_levels(&self) -> &Array1<f64>

Get noise levels

Examples found in repository

examples/quantum_diffusion.rs (line 185)

fn score_diffusion_demo() -> Result<()> {
    // Create score-based model
    let model = QuantumScoreDiffusion::new(
        2,  // data dimension
        4,  // num qubits
        10, // noise levels
    )?;

    println!("   Created quantum score-based diffusion model");
    println!("   - Noise levels: {:?}", model.noise_levels());

    // Test score estimation
    let x = Array1::from_vec(vec![0.5, -0.3]);
    let noise_level = 0.1;

    let score = model.estimate_score(&x, noise_level)?;
    println!("\n   Score estimation:");
    println!("   - Input: [{:.3}, {:.3}]", x[0], x[1]);
    println!("   - Noise level: {noise_level:.3}");
    println!("   - Estimated score: [{:.3}, {:.3}]", score[0], score[1]);

    // Langevin sampling
    println!("\n   Langevin sampling:");
    let init = Array1::from_vec(vec![2.0, 2.0]);
    let num_steps = 100;
    let step_size = 0.01;

    let sample = model.langevin_sample(init.clone(), noise_level, num_steps, step_size)?;

    println!("   - Initial: [{:.3}, {:.3}]", init[0], init[1]);
    println!(
        "   - After {} steps: [{:.3}, {:.3}]",
        num_steps, sample[0], sample[1]
    );
    println!(
        "   - Distance moved: {:.3}",
        (sample[0] - init[0]).hypot(sample[1] - init[1])
    );

    Ok(())
}