scHopfield.dynamics.ODESolver

class scHopfield.dynamics.ODESolver(W: ndarray, bias_vector: ndarray, gamma: ndarray, threshold: ndarray, exponent: ndarray, x_min: float = 0.0, x_max: ndarray | None = None, fixed_indices: ndarray | None = None, fixed_values: ndarray | None = None)[source]

ODE solver for Hopfield network dynamics.

Solves: dx/dt = W * sigmoid(x) - gamma * x + I

With constraints to ensure expression values remain non-negative and bounded to prevent divergence. Supports fixing certain genes at constant values (e.g., for knockout/overexpression simulations).

__init__(W: ndarray, bias_vector: ndarray, gamma: ndarray, threshold: ndarray, exponent: ndarray, x_min: float = 0.0, x_max: ndarray | None = None, fixed_indices: ndarray | None = None, fixed_values: ndarray | None = None)[source]

Initialize ODE solver.

Parameters:

  • W (np.ndarray) – Interaction matrix

  • I (np.ndarray) – External input

  • gamma (np.ndarray) – Degradation rates

  • threshold (np.ndarray) – Sigmoid threshold parameters

  • exponent (np.ndarray) – Sigmoid exponent parameters

  • x_min (float, optional (default: 0.0)) – Minimum expression value (non-negative constraint)

  • x_max (np.ndarray, optional) – Maximum expression values per gene. If None, no upper bound.

  • fixed_indices (np.ndarray, optional) – Indices of genes to keep fixed (e.g., perturbed genes)

  • fixed_values (np.ndarray, optional) – Values to fix the genes at (must match length of fixed_indices)

Methods

__init__(W, bias_vector, gamma, threshold, ...)

Initialize ODE solver.

dynamics(x, t)

Compute dx/dt with soft boundary enforcement.

dynamics_batch(X, t)

Compute dx/dt for a batch of states (n_cells, n_genes).

dynamics_ivp(t, x)

Compute dx/dt for solve_ivp (arguments reversed).

set_fixed_genes(fixed_indices, fixed_values)

Set genes to be held fixed during simulation.

solve(x0, t_span[, method, clip_each_step])

Solve ODE from initial condition x0.