Source code for upxo.algorithms.alg201

from copy import deepcopy
from random import sample as sample_rand
import numpy as np
import numpy.random as rand
from upxo.pxtal.mcgs2_temporal_slice import mcgs2_grain_structure as GS2d
import random
from numba import njit

@njit
def unique_with_counts(arr):
    """
    This is because, current numba is incompatible with the return_counts
    argument in np.unique.
    """
    # Sort the array for grouping identical values
    sorted_arr = np.sort(arr)
    unique_vals = []
    counts = []

    # Initialize the first unique value and count
    if len(sorted_arr) > 0:
        current_val = sorted_arr[0]
        count = 1

        for i in range(1, len(sorted_arr)):
            if sorted_arr[i] == current_val:
                count += 1
            else:
                unique_vals.append(current_val)
                counts.append(count)
                current_val = sorted_arr[i]
                count = 1

        # Append the last value and count
        unique_vals.append(current_val)
        counts.append(count)

    return np.array(unique_vals), np.array(counts)

@njit
def mcloop_alg201(cbp, sbp, S, AIA0, AIA1, NLM, rsfso):
    """Mcloop alg201."""
    # rsfso = 2
    for s0 in list(range(S.shape[0])):  # along axis 0
        s00, s01, s02 = s0+0, s0+1, s0+2
        for s1 in list(range(S.shape[1])):  # along axis 1
            s10, s11, s12 = s1+0, s1+1, s1+2
            ssub_00 = S[AIA0[s00, s10], AIA1[s00, s10]]
            ssub_01 = S[AIA0[s01, s10], AIA1[s01, s10]]
            ssub_02 = S[AIA0[s02, s10], AIA1[s02, s10]]
            ssub_10 = S[AIA0[s00, s11], AIA1[s00, s11]]
            ssub_11 = S[AIA0[s01, s11], AIA1[s01, s11]]
            ssub_12 = S[AIA0[s02, s11], AIA1[s02, s11]]
            ssub_20 = S[AIA0[s00, s12], AIA1[s00, s12]]
            ssub_21 = S[AIA0[s01, s12], AIA1[s01, s12]]
            ssub_22 = S[AIA0[s02, s12], AIA1[s02, s12]]
            Neigh = np.array([ssub_00, ssub_01, ssub_02,
                              ssub_10, ssub_11, ssub_12,
                              ssub_20, ssub_21, ssub_22])
            if Neigh.min() != Neigh.max():
                DelH1 = NLM[0, 0]*int(ssub_11 == ssub_00) + \
                        NLM[0, 1]*int(ssub_11 == ssub_01) + \
                        NLM[0, 2]*int(ssub_11 == ssub_02) + \
                        NLM[1, 0]*int(ssub_11 == ssub_10) + \
                        NLM[1, 2]*int(ssub_11 == ssub_12) + \
                        NLM[2, 0]*int(ssub_11 == ssub_20) + \
                        NLM[2, 1]*int(ssub_11 == ssub_21) + \
                        NLM[2, 2]*int(ssub_11 == ssub_22)
                # ---------------------------------------------
                # If the sampling is to be selected with
                # weightage to dominant neighbour state, then:
                Neigh = Neigh[Neigh != ssub_11]
                # ---------------------------------------------
                # Neigh_, counts_ = np.unique(Neigh, return_counts=True)
                Neigh_, counts_ = unique_with_counts(Neigh)

                counts_ = np.argsort(counts_)[::-1]
                counts_ = counts_[:rsfso]

                random_index = np.random.randint(0, len(counts_))
                ssub_11_b = Neigh_[random_index]
                # ---------------------------------------------
                DelH2 = NLM[0, 0]*int(ssub_11_b == ssub_00) + \
                        NLM[0, 1]*int(ssub_11_b == ssub_01) + \
                        NLM[0, 2]*int(ssub_11_b == ssub_02) + \
                        NLM[1, 0]*int(ssub_11_b == ssub_10) + \
                        NLM[1, 2]*int(ssub_11_b == ssub_12) + \
                        NLM[2, 0]*int(ssub_11_b == ssub_20) + \
                        NLM[2, 1]*int(ssub_11_b == ssub_21) + \
                        NLM[2, 2]*int(ssub_11_b == ssub_22)
                if DelH2 >= DelH1:
                    S[s0, s1] = ssub_11_b
                elif cbp:
                    if sbp[int(ssub_11_b-1)] < np.random.random():
                        S[s0, s1] = ssub_11_b
    return S

[docs] def run(uisim, uiint, uidata, uigrid, rsfso, xgr, ygr, zgr, px_size, _a, _b, _c, S, AIA0, AIA1, display_messages): """Run.""" # --------------------------------------------- print("Using ALG-200: SA's NL-1 weighted Q-Pott's model:") print('|' + 15*'-'+' MC SIM RUN IN PROGRESS on: ALG201' + 15*'-' + '|') gs = {} # Build the Non-Locality Matrix NLM_00, NLM_01, NLM_02 = _a # Unpack 3 colms of 1st row NLM_10, NLM_11, NLM_12 = _b # Unpack 3 colms of 2nd row NLM_20, NLM_21, NLM_22 = _c # Unpack 3 colms of 3rd row NLM = np.array([[NLM_00, NLM_01, NLM_02], [NLM_10, NLM_11, NLM_12], [NLM_20, NLM_21, NLM_22] ]) print('============================================') print(NLM) print('============================================') # --------------------------------------------- # Begin modified Markov-Chain annealing iterations fully_annealed = False fully_annealed_at_m = None for m in range(uisim.mcsteps): if S.min() == S.max(): print(30*'.') print(f'Single crystal achieved at iteration {m}.') fully_annealed, fully_annealed_at_m = True, m # Store the last temporal slice as a UPXO grain structure by # default gs[m] = GS2d(m=m, dim=uigrid.dim, uidata=uidata, px_size=px_size, S_total=uisim.S, xgr=xgr, ygr=ygr, uigrid=uigrid, ) gs[m].s = deepcopy(S) if display_messages: print(f"GS temporal slice {m} stored\n\n" "!! MONTE-CARLO ALG.202 run ended !!\n") # 3952493 break else: cbp = uisim.consider_boltzmann_probability sbp = uisim.s_boltz_prob S = mcloop_alg201(cbp, sbp, S, AIA0, AIA1, NLM, rsfso) cond_1 = m % uiint.mcint_save_at_mcstep_interval == 0.0 save_msg = False if m==0 or cond_1 or fully_annealed: gs[m] = GS2d(m=m, dim=uigrid.dim, uidata=uidata, px_size=px_size, S_total=uisim.S, xgr=xgr, ygr=ygr, uigrid=uigrid, ) gs[m].s = deepcopy(S) save_msg = True if display_messages: print(f"GS temporal slice {m} stored") if m % uiint.mcint_promt_display == 0: if display_messages: if not save_msg: print(f"Monte-Carlo temporal step = {m}") print('|' + 15*'-'+' MC SIM RUN COMPLETED on: ALG201' + 15*'-' + '|') fully_annealed = {'fully_annealed': fully_annealed, 'm': fully_annealed_at_m} return gs, fully_annealed