Struct SolverState

Source

pub struct SolverState {
    pub assignments: Vec<f32>,
    pub velocities: Vec<f32>,
    pub gradients: Vec<f32>,
}

Expand description

Internal state during CLS optimization.

Tracks the continuous variable assignments, momentum velocities, and computed gradients throughout the optimization process.

§Memory Layout

Each vector has length equal to the number of variables:

assignments: Current continuous values in [0,1]
velocities: Momentum velocities (accumulated gradient history)
gradients: Computed gradients for the current iteration

Fields§

§assignments: Vec<f32>

Continuous variable assignments in [0,1].

Index i corresponds to variable i. Values close to 1.0 indicate the variable should be true, values close to 0.0 indicate false.

§velocities: Vec<f32>

Momentum velocities for each variable.

Accumulates gradient history to help escape local minima and smooth the optimization trajectory.

§gradients: Vec<f32>

Computed gradients for the current iteration.

The gradient of the unsatisfied clause penalty with respect to each variable. Negative gradients indicate the variable should increase to reduce unsatisfaction.

Implementations§

Source §

impl SolverState

Source

pub fn new(num_vars: u32) -> Self

Creates a new solver state for the given number of variables.

Variables are initialized with values near 0.5 using a deterministic pseudo-random pattern based on index to break symmetry while maintaining reproducibility.

§Arguments

num_vars - Number of variables in the SAT instance

§Example

use xlog_solve::SolverState;

let state = SolverState::new(10);
assert_eq!(state.assignments.len(), 10);
assert_eq!(state.velocities.len(), 10);
assert_eq!(state.gradients.len(), 10);

Source

pub fn with_assignments(assignments: Vec<f32>) -> Self

Creates a solver state with specific initial assignments.

Useful for warm-starting from a known assignment or for testing.

§Arguments

assignments - Initial continuous assignments in [0,1]

Source

pub fn discretize(&self, threshold: f32) -> Vec<bool>

Discretizes the continuous assignments to boolean values.

Values >= threshold become true, values < threshold become false.

§Arguments

threshold - The threshold for boolean conversion (typically 0.5)

§Returns

A vector of boolean values, one per variable.

§Example

use xlog_solve::SolverState;

let mut state = SolverState::new(3);
state.assignments = vec![0.3, 0.7, 0.5];
let discrete = state.discretize(0.5);
assert_eq!(discrete, vec![false, true, true]);