Fuzzy fitting

In [1]:

import numpy as np

from example_models import get_upper_glycolysis
from mxlpy import Simulator, plot
from mxlpy.fuzzy import ThompsonState, thompson_sampling

# Create artificial data
p_orig = {
    "k1": 0.25,
    "k2": 1.2,
    "k3": 1.6,
    "k3m": 0.5,
    "k4": 100,
    "k5": 1.3,
    "k6": 1.5,
    "k7": 2.5,
}
m = get_upper_glycolysis().update_parameters(p_orig)
data, _ = (
    Simulator(m)
    .simulate_time_course(np.linspace(0, 10, 11, dtype=float))
    .get_result()
    .unwrap_or_err()
)

# Perform sampling
state = thompson_sampling(
    m,
    data=data,
    state=ThompsonState.from_parameter_values(
        {
            "k1": np.geomspace(1e-1, 1e2, 21),
            "k2": np.geomspace(1e-1, 1e2, 21),
            "k3": np.geomspace(1e-1, 1e2, 21),
            "k3m": np.geomspace(1e-1, 1e2, 21),
            "k4": np.geomspace(1e-1, 1e2, 21),
            "k5": np.geomspace(1e-1, 1e2, 21),
            "k6": np.geomspace(1e-1, 1e2, 21),
            "k7": np.geomspace(1e-1, 1e2, 21),
        }
    ),
    rtol=0.11,
    n=1000,
    parallel=True,
)

import numpy as np from example_models import get_upper_glycolysis from mxlpy import Simulator, plot from mxlpy.fuzzy import ThompsonState, thompson_sampling # Create artificial data p_orig = { "k1": 0.25, "k2": 1.2, "k3": 1.6, "k3m": 0.5, "k4": 100, "k5": 1.3, "k6": 1.5, "k7": 2.5, } m = get_upper_glycolysis().update_parameters(p_orig) data, _ = ( Simulator(m) .simulate_time_course(np.linspace(0, 10, 11, dtype=float)) .get_result() .unwrap_or_err() ) # Perform sampling state = thompson_sampling( m, data=data, state=ThompsonState.from_parameter_values( { "k1": np.geomspace(1e-1, 1e2, 21), "k2": np.geomspace(1e-1, 1e2, 21), "k3": np.geomspace(1e-1, 1e2, 21), "k3m": np.geomspace(1e-1, 1e2, 21), "k4": np.geomspace(1e-1, 1e2, 21), "k5": np.geomspace(1e-1, 1e2, 21), "k6": np.geomspace(1e-1, 1e2, 21), "k7": np.geomspace(1e-1, 1e2, 21), } ), rtol=0.11, n=1000, parallel=True, )

  0%|          | 0/1000 [00:00<?, ?it/s]

  0%|          | 1/1000 [00:03<1:05:01,  3.91s/it]

  1%|          | 8/1000 [00:04<06:05,  2.72it/s]  

  1%|▏         | 13/1000 [00:04<03:34,  4.61it/s]

  2%|▏         | 21/1000 [00:04<02:12,  7.39it/s]

  6%|▌         | 59/1000 [00:04<00:29, 31.45it/s]

  9%|▉         | 92/1000 [00:04<00:16, 56.65it/s]

 13%|█▎        | 133/1000 [00:05<00:09, 94.16it/s]

 18%|█▊        | 177/1000 [00:05<00:05, 138.57it/s]

 21%|██        | 209/1000 [00:05<00:05, 154.37it/s]

 26%|██▌       | 256/1000 [00:05<00:03, 208.77it/s]

 30%|███       | 301/1000 [00:05<00:02, 255.97it/s]

 34%|███▍      | 339/1000 [00:05<00:02, 279.82it/s]

 38%|███▊      | 382/1000 [00:05<00:01, 315.01it/s]

 42%|████▏     | 422/1000 [00:05<00:01, 335.93it/s]

 47%|████▋     | 466/1000 [00:05<00:01, 362.71it/s]

 51%|█████     | 512/1000 [00:06<00:01, 388.68it/s]

 55%|█████▌    | 554/1000 [00:06<00:01, 394.79it/s]

 60%|█████▉    | 598/1000 [00:06<00:00, 406.44it/s]

 64%|██████▍   | 645/1000 [00:06<00:00, 422.32it/s]

 69%|██████▉   | 689/1000 [00:06<00:00, 421.64it/s]

 73%|███████▎  | 732/1000 [00:06<00:00, 421.55it/s]

 78%|███████▊  | 778/1000 [00:06<00:00, 432.10it/s]

 82%|████████▏ | 822/1000 [00:06<00:00, 430.59it/s]

 87%|████████▋ | 872/1000 [00:06<00:00, 449.51it/s]

 92%|█████████▏| 918/1000 [00:06<00:00, 448.05it/s]

 97%|█████████▋| 969/1000 [00:07<00:00, 447.74it/s]

1012it [00:07, 141.38it/s]                         

In [2]:

from scipy.interpolate import CubicSpline

# Visualize results
fig, axs = plot.grid_layout(len(state.state), n_cols=4)
ai = iter(axs)

for ax, (name, var) in zip(axs, state.state.items(), strict=False):
    ax.set(title=f"{name}")
    xticks = np.arange(len(var["x"]))

    ax.bar(xticks, var["success"] / var["fail"])
    ax.set_xticks(xticks)
    ax.set_xticklabels((f"{i:.0e}" for i in var["x"]), rotation=90)

    x_trans = CubicSpline(var["x"], xticks)
    x_data = x_trans(p_orig[name])
    y_data = ax.transData.inverted().transform(ax.transAxes.transform([0, 0.5]))[1]
    ax.annotate(
        p_orig[name],
        xy=(x_data, 0),  # type: ignore
        xytext=(x_data, y_data),  # type: ignore
        ha="center",
        arrowprops={"arrowstyle": "->"},
    )

from scipy.interpolate import CubicSpline # Visualize results fig, axs = plot.grid_layout(len(state.state), n_cols=4) ai = iter(axs) for ax, (name, var) in zip(axs, state.state.items(), strict=False): ax.set(title=f"{name}") xticks = np.arange(len(var["x"])) ax.bar(xticks, var["success"] / var["fail"]) ax.set_xticks(xticks) ax.set_xticklabels((f"{i:.0e}" for i in var["x"]), rotation=90) x_trans = CubicSpline(var["x"], xticks) x_data = x_trans(p_orig[name]) y_data = ax.transData.inverted().transform(ax.transAxes.transform([0, 0.5]))[1] ax.annotate( p_orig[name], xy=(x_data, 0), # type: ignore xytext=(x_data, y_data), # type: ignore ha="center", arrowprops={"arrowstyle": "->"}, )

In [ ]:

In [ ]: