SEC.Models.SdmOptimizer

SEC.Models.SdmOptimizer#

SEC.Models.SdmOptimizer.py

optimize_sdm_xr_decomposition(decomposition, env_params, model_params=None, **kwargs)#

Optimize the SDM decomposition.

Parameters:

  • decomposition (Decomposition) – The decomposition to optimize.

  • env_params (tuple) – The environmental parameters (N, T, me, mp, N0, t0, poresize).

  • model_params (dict, optional) – The parameters for the SDM model.

  • kwargs (dict) – Additional parameters for the optimization process.

Returns:

new_xr_ccurves – The optimized SDM component curves.

Return type:

list of SdmComponentCurve

refine_sdm_per_component(sdm_ccurves, x, y, **kwargs)#

Pass 2 (lighter): fix shared column params, refine per-component (rg, scale) only.

This is the SDM analog of refine_edm_per_component. After Pass 1 has established a good shared-column characterisation (N, T, me, mp, x0, tI, N0, poresize, timescale, k), this pass holds those fixed and adjusts only the per-component radius-of-gyration and amplitude — a lower-dimensional problem that is less prone to degeneracy under high component overlap.

Parameters:

  • sdm_ccurves (list of SdmComponentCurve) – Pass-1 optimised SDM component curves (mono or lognormal pore_dist).

  • x (array) – Frame positions (same grid as used by the component curves).

  • y (array) – Total XR elution curve values to fit.

Returns:

Refined curves with fixed shared column and optimised per-component (rg, scale). The column object is shared across all returned curves.

Return type:

list of SdmComponentCurve

optimize_sdm_uv_decomposition(decomposition, xr_ccurves, **kwargs)#

Optimize the SDM UV decomposition.

Parameters:

  • decomposition (Decomposition) – The decomposition to optimize.

  • xr_ccurves (list of SdmComponentCurve) – The SDM component curves from the XR decomposition.

  • kwargs (dict) – Additional parameters for the optimization process.

Returns:

new_uv_ccurves – The optimized UV component curves.

Return type:

list of UvComponentCurve

optimize_sdm_lognormal_xr_decomposition(decomposition, env_params, model_params=None, **kwargs)#

Optimize XR decomposition with SDM lognormal pore distribution.

Parameters:

  • decomposition (Decomposition) – The decomposition to optimize.

  • env_params (tuple) – The environmental parameters (N, T, me, mp, N0, t0, mu, sigma).

  • model_params (dict, optional) –

    The parameters for the SDM model. Recognized keys:

    ln_pore_sigmafloat or None (default 0.3)

    Standard deviation of ln(r/r0), held constant during optimization (removed from the free-parameter vector). Invariant to the choice of length unit (a unit change only shifts all ln(r/r0) values by a constant, which drops out of the standard deviation). The default 0.3 implies pores span a factor of exp(0.3) ≈ 1.35 around the median — a reasonable prior for most SEC columns. Pass None to make sigma a free parameter; note that sigma is underdetermined when num_components=1 and will drift to sigma_max if left free.

    sigma_maxfloat (default 0.8)

    Upper bound on sigma when it is free.

    k : float (default 2.0) rt_dist : str (default 'gamma')

  • kwargs (dict) – Additional parameters for the optimization process.

Returns:

new_xr_ccurves – The optimized SDM component curves with lognormal pore distribution.

Return type:

list of SdmComponentCurve

adjust_rg_and_poresize(sdm_decomposition)#

Adjust rg and poresize in the decomposition based on the optimized component curves.