Corn NIR benchmark (Cargill / Eigenvector)

This notebook walks through SpectoPrep on the public corn NIR dataset (m5 instrument): raw spectra, nested cross-validated optimisation, consensus preprocessing per constituent, and predicted-versus-reference plots.

Constituents: Moisture, Oil, Protein, Starch.

Full paper-grade nested CV (init_points=10, n_iter=50, i.e. 60 evaluations × 5 outer folds × 4 constituents) takes several minutes per seed. The paper reports the mean ± SD over ten seeds (0–9); examples/run_corn_benchmark.py runs one seed (seed 0) and examples/corn_results/multiseed_summary.json holds the ten-seed summary. Prefer loading these precomputed results, or use the shorter live demo cells below.

1. Load data

from pathlib import Path
import json
import numpy as np
import scipy.io as sio
import matplotlib.pyplot as plt

from spectoprep import PipelineOptimizer, SpectroPrepPlotter

COMPONENTS = ["Moisture", "Oil", "Protein", "Starch"]

# Resolve repo examples/ whether this notebook is under docs/notebooks/ or notebook/
for candidate in [Path("../../examples"), Path("../examples"), Path("examples")]:
    if candidate.exists():
        EXAMPLES = candidate.resolve()
        break
else:
    EXAMPLES = Path("examples").resolve()

CORN_CANDIDATES = [
    EXAMPLES / "data" / "corn.mat",
    Path("/tmp/spectoprep-data/corn.mat"),
]
CORN = next((p for p in CORN_CANDIDATES if p.exists()), None)
assert CORN is not None, (
    "Corn dataset not found. Place corn.mat at examples/data/corn.mat "
    "or /tmp/spectoprep-data/corn.mat (https://eigenvector.com/data/Corn/)."
)

d = sio.loadmat(CORN, struct_as_record=False, squeeze_me=True)
X = np.asarray(d["m5spec"].data, dtype=float)
Y = np.asarray(d["propvals"].data, dtype=float)
try:
    wavelengths = np.asarray(d["m5spec"].axisscale[1], dtype=float).ravel()
    if wavelengths.size != X.shape[1]:
        raise ValueError
except Exception:
    wavelengths = np.linspace(1100, 2498, X.shape[1])

print(f"Loaded {CORN}")
print(f"X={X.shape}, Y={Y.shape}, λ=[{wavelengths.min():.0f}, {wavelengths.max():.0f}] nm")
print("Constituents:", COMPONENTS)

Loaded /Users/habeebbabatunde/Downloads/Spectoprep-main/examples/data/corn.mat
X=(80, 700), Y=(80, 4), λ=[1100, 2498] nm
Constituents: ['Moisture', 'Oil', 'Protein', 'Starch']

2. Raw spectra (full set, inverted wavelength axis)

SpectroPrepPlotter.set_style(context="notebook", font_scale=1.1)
SpectroPrepPlotter.plot_spectra(
    wavelengths,
    X,
    title="Corn NIR (m5) — raw spectra (n=80)",
    xlabel="Wavelength (nm)",
    ylabel="Absorbance",
    alpha=0.35,
    invert_xaxis=True,
)
plt.show()

3. Load tracked optimal preprocessing (from full benchmark)

Run once from the repo root:

python examples/run_corn_benchmark.py

This writes examples/corn_results/optimal_pipelines.json with per-fold chains, consensus pipelines, and nested-CV metrics for every constituent. These are the seed-0 results; the ten-seed mean ± SD behind the paper’s tables is shown right after.

results_path = EXAMPLES / "corn_results" / "optimal_pipelines.json"
if results_path.exists():
    summary = json.loads(results_path.read_text())
    print("Protocol:", json.dumps(summary["protocol"], indent=2))
    print("\nConsensus optimal preprocessing per constituent:")
    for name, block in summary["constituents"].items():
        chain = " → ".join(block["consensus_pipeline"])
        spx = block["spectoprep"]
        print(f"  {name:9s}: {chain}")
        print(f"             R²={spx['r2']:.3f}, RMSE={spx['rmse']:.4f}, RPD={spx['rpd']:.1f}")
        print(f"             fold chains: {block['fold_pipelines']}")
else:
    summary = None
    print(f"No precomputed results at {results_path}")
    print("Run: python examples/run_corn_benchmark.py")

Protocol: {
  "instrument": "m5",
  "outer_cv": "KFold(n_splits=5, shuffle=True, random_state=0)",
  "inner_cv": "group_kfold n_splits=3",
  "init_points": 10,
  "n_iter": 50,
  "evaluations_per_outer_fold": 60,
  "preprocessing_steps": [
    "snv",
    "msc",
    "savgol",
    "detrend",
    "scaler",
    "robust_scaler",
    "meancn"
  ]
}

Consensus optimal preprocessing per constituent:
  Moisture : scaler → meancn
             R²=1.000, RMSE=0.0077, RPD=49.6
             fold chains: [['scaler', 'meancn'], ['scaler', 'meancn'], ['scaler'], ['scaler', 'meancn'], ['scaler', 'meancn']]
  Oil      : detrend → scaler
             R²=0.973, RMSE=0.0291, RPD=6.1
             fold chains: [['detrend', 'scaler'], ['scaler', 'meancn'], ['detrend', 'robust_scaler'], ['detrend', 'scaler'], ['detrend', 'scaler']]
  Protein  : detrend → robust_scaler
             R²=0.990, RMSE=0.0488, RPD=10.2
             fold chains: [['detrend', 'robust_scaler'], ['detrend', 'scaler'], ['detrend', 'robust_scaler'], ['detrend', 'robust_scaler'], ['detrend', 'robust_scaler']]
  Starch   : scaler
             R²=0.980, RMSE=0.1167, RPD=7.0
             fold chains: [['scaler'], ['scaler'], ['savgol', 'scaler'], ['scaler', 'meancn'], ['msc', 'robust_scaler']]

Ten-seed summary (paper tables)

Mean ± SD over seeds of the pooled out-of-fold metrics, from examples/corn_results/multiseed_summary.json.

multi_path = EXAMPLES / "corn_results" / "multiseed_summary.json"
if multi_path.exists():
    multi = json.loads(multi_path.read_text())
    proto = multi["protocol"]
    print(f"Seeds {proto['seeds'][0]}-{proto['seeds'][-1]} | init_points={proto['init_points']}, "
          f"n_iter={proto['n_iter']} | ridge alpha grid {proto['ridge_alpha_grid']}")
    print(f"{'Constituent':12s} {'Raw RMSE':>15s} {'SpectoPrep RMSE':>19s} {'SpectoPrep R²':>16s} {'RPD':>11s}")
    for name, block in multi["constituents"].items():
        raw, spx = block["raw"], block["spectoprep"]
        print(
            f"{name:12s} {raw['rmse']['mean']:7.3f} ± {raw['rmse']['sd']:.3f}"
            f"   {spx['rmse']['mean']:8.4f} ± {spx['rmse']['sd']:.4f}"
            f"   {spx['r2']['mean']:.3f} ± {spx['r2']['sd']:.3f}"
            f"   {spx['rpd']['mean']:5.1f} ± {spx['rpd']['sd']:.1f}"
        )
else:
    print(f"No ten-seed summary at {multi_path}")

Seeds 0-9 | init_points=10, n_iter=50 | ridge alpha grid logspace(-3, 3, 25)
Constituent         Raw RMSE     SpectoPrep RMSE    SpectoPrep R²         RPD
Moisture       0.046 ± 0.001     0.0080 ± 0.0008   1.000 ± 0.000    47.8 ± 4.9
Oil            0.078 ± 0.002     0.0303 ± 0.0050   0.970 ± 0.010     6.0 ± 1.0
Protein        0.180 ± 0.009     0.0627 ± 0.0133   0.983 ± 0.007     8.3 ± 2.1
Starch         0.408 ± 0.021     0.1080 ± 0.0175   0.982 ± 0.005     7.8 ± 1.4

4. Short live demo on one constituent (Protein)

A reduced search for interactive use. For paper numbers use the precomputed nested-CV results.

component = "Protein"
y = Y[:, COMPONENTS.index(component)]
groups = np.arange(len(y))

# Hold out 25% for a quick illustration (not the nested-CV paper protocol)
from sklearn.model_selection import GroupShuffleSplit
tr, te = next(GroupShuffleSplit(n_splits=1, test_size=0.25, random_state=21).split(X, y, groups))

opt = PipelineOptimizer(
    X_train=X[tr], y_train=y[tr],
    X_test=X[te], y_test=y[te],
    groups=groups[tr],
    preprocessing_steps=["snv", "msc", "savgol", "detrend", "scaler", "robust_scaler", "meancn"],
    cv_method="group_kfold",
    n_splits=3,
    max_pipeline_length=2,
    allowed_preprocess_combinations=[1, 2],
    random_state=42,
    log_level="WARNING",
)
best_params, best_pipeline = opt.bayesian_optimize(init_points=5, n_iter=15)
preds, rmse, r2 = opt.get_best_pipeline_predictions(best_pipeline)
print("Selected:", " → ".join(s for s, _ in best_pipeline.steps[:-1]))
print(f"Held-out {component}: RMSE={rmse:.4f}, R²={r2:.4f}")

|   iter    |  target   |  als_lam  |   als_p   | als_niter | savgol... | savgol... | savgol... | detren... | detren... | emsc_o... | global... | lsnv_win  |  rnv_lp   |  rnv_up   | pca_n_... | rs_qua... | rs_qua... | rs_wit... | rs_wit... | minmax... | minmax... | power_... | power_... | quanti... | quanti... | ica_n_... | kpca_n... | kpca_k... | kpca_g... | lle_n_... | lle_n_... |   skb_k   | fa_n_c... | fa_lin... | rbf_n_... | rbf_gamma | pipeli... |
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| 1         | -0.201306 | 37516.557 | 0.2857071 | 39.279757 | 10.591950 | 1.6240745 | 0.4679835 | 0.0580836 | 4.4647045 | 3.4044600 | 7.3726532 | 7.1235069 | 19.548647 | 90.811066 | 22.809232 | 7.7273745 | 74.585112 | 0.3042422 | 0.5247564 | 0.1727780 | 0.7164916 | 0.6118528 | 0.1394938 | 65.507483 | 0.3663618 | 23.891359 | 39.688446 | 0.7966983 | 5.1428301 | 18.587607 | 6.1612603 | 61.539395 | 18.711364 | 0.1945042 | 190.28825 | 0.9656664 | 20.209933 |
| 2         | -0.112765 | 30530.915 | 0.0383249 | 37.369321 | 9.6409149 | 1.4881529 | 1.4855307 | 0.0343885 | 4.6372816 | 2.0351199 | 6.9627005 | 8.8702664 | 12.801020 | 83.667756 | 20.115736 | 19.543769 | 89.378320 | 0.9394989 | 0.8948273 | 0.2391599 | 0.9687496 | 0.0884925 | 0.1959828 | 18.593184 | 0.3253303 | 20.656509 | 15.024753 | 3.3066626 | 3.5681765 | 9.8661662 | 18.567402 | 15.810574 | 80.615304 | 0.2229064 | 197.50851 | 0.7724725 | 4.9678920 |
| 3         | -0.201267 | 651.65950 | 0.2464838 | 38.274293 | 11.374043 | 4.0850813 | 0.2221339 | 0.3584657 | 1.4634762 | 4.4524137 | 6.6096831 | 8.9853881 | 5.9533752 | 77.774558 | 33.867965 | 15.944092 | 85.938936 | 0.8872127 | 0.4722149 | 0.0478376 | 0.8852979 | 0.7607850 | 0.5612771 | 156.48376 | 0.4937955 | 27.091175 | 22.521968 | 0.1014223 | 1.0798063 | 2.8800171 | 20.910260 | 32.806886 | 51.839927 | 2.7136237 | 57.365523 | 0.4109725 | 18.888778 |
| 4         | -0.135801 | 22956.936 | 0.0323241 | 21.590058 | 7.9673277 | 4.7187906 | 2.4243611 | 0.6334037 | 4.4858423 | 4.2146883 | 2.6791305 | 12.355353 | 13.090133 | 90.186003 | 89.816947 | 9.7700521 | 72.751298 | 0.2279351 | 0.4271077 | 0.3272059 | 0.9442922 | 0.0069521 | 0.5107473 | 89.308090 | 0.2221078 | 7.7535376 | 18.205528 | 3.7622097 | 3.2327061 | 16.526137 | 22.575473 | 37.635701 | 97.234644 | 2.8777174 | 57.838636 | 0.4977512 | 7.5219577 |
| 5         | -0.199573 | 28555.565 | 0.0206972 | 34.382573 | 10.016074 | 1.2059150 | 0.8359393 | 0.9082658 | 1.9582475 | 1.5795794 | 5.4050748 | 12.913902 | 8.6308290 | 86.803388 | 76.638722 | 8.5645631 | 88.205408 | 0.3677831 | 0.6323058 | 0.2534118 | 0.8143098 | 0.0902897 | 0.8353024 | 70.948212 | 0.1865185 | 3.9572067 | 30.362861 | 2.7034818 | 0.1668617 | 16.338605 | 10.662394 | 65.226933 | 19.087910 | 2.0659038 | 83.479715 | 0.9367932 | 3.4380236 |
| 6         | -0.135147 | 66032.406 | 0.1534903 | 19.372764 | 9.3918661 | 4.8684740 | 2.1052207 | 0.8475333 | 4.0728043 | 1.3122712 | 8.4906219 | 11.316496 | 14.718317 | 75.721621 | 22.234906 | 12.681580 | 86.114857 | 0.0073249 | 0.9438948 | 0.1130865 | 0.7581971 | 0.8603996 | 0.4619540 | 129.39196 | 0.9200338 | 40.379079 | 41.101876 | 3.9245586 | 7.4256003 | 9.3097008 | 25.137096 | 4.1787315 | 89.810840 | 1.7464904 | 91.333266 | 0.8952873 | 11.393923 |
| 7         | -0.486287 | 70264.575 | 0.1463862 | 16.222442 | 8.2036242 | 1.5677439 | 1.3177253 | 0.1534522 | 2.2505267 | 2.5206960 | 8.3861071 | 8.5491592 | 5.7405904 | 81.743264 | 84.070102 | 10.661843 | 77.253459 | 0.4770757 | 0.3484341 | 0.2824022 | 0.6217481 | 0.0627184 | 0.7140648 | 101.59952 | 0.2117297 | 20.560554 | 48.555231 | 1.4943151 | 1.1619394 | 10.747585 | 18.474039 | 75.535467 | 11.910267 | 1.0707840 | 71.585879 | 0.4952237 | 19.129795 |
| 8         | -0.201291 | 32109.429 | 0.2503331 | 12.893324 | 7.4250740 | 3.1372413 | 2.1633247 | 0.2021667 | 2.0202262 | 3.6082898 | 8.9409638 | 10.561138 | 11.299048 | 87.267648 | 28.601137 | 11.479725 | 94.641740 | 0.5560556 | 0.3737704 | 0.2743326 | 0.9171885 | 0.2223209 | 0.9548093 | 166.87569 | 0.1832917 | 19.191693 | 28.894459 | 2.4855787 | 0.4512550 | 22.972988 | 21.639186 | 41.986897 | 32.568161 | 0.7361100 | 121.21571 | 0.5704438 | 6.0884562 |
| 9         | -0.203396 | 64109.709 | 0.2932527 | 35.855925 | 10.656316 | 2.1951431 | 1.5633786 | 0.2060850 | 2.5087700 | 1.9700171 | 7.6154658 | 11.209897 | 18.255611 | 78.178874 | 56.709866 | 7.4454476 | 71.575709 | 0.3185483 | 0.2787127 | 0.3573972 | 0.9514880 | 0.5654358 | 0.8282916 | 29.720157 | 0.6521224 | 25.573632 | 32.199097 | 0.3202625 | 3.8215256 | 8.2847733 | 15.270145 | 24.464196 | 78.476695 | 0.6928583 | 96.493887 | 0.9319509 | 21.729746 |
| 10        | -0.215379 | 56416.745 | 0.1427190 | 12.247134 | 11.290807 | 4.2649308 | 0.7609171 | 0.8934749 | 2.5122530 | 4.5036438 | 1.6198449 | 9.5859924 | 18.282919 | 73.290938 | 51.599543 | 19.068249 | 79.782782 | 0.8988935 | 0.3828247 | 0.0667432 | 0.8572721 | 0.8790709 | 0.2592703 | 18.077940 | 0.6717897 | 22.738390 | 29.045481 | 0.1734994 | 3.6088029 | 21.934029 | 12.484418 | 95.898601 | 38.164928 | 1.2795519 | 194.77494 | 0.1065600 | 15.168963 |
| 11        | -0.135147 | 21120.227 | 0.2637671 | 33.670373 | 7.6418389 | 3.8554569 | 0.3103593 | 0.7112061 | 4.3879975 | 4.3442871 | 1.7018817 | 11.399295 | 19.986704 | 91.955049 | 64.516415 | 15.871078 | 87.837685 | 0.1871057 | 0.1048155 | 0.2298309 | 0.6025569 | 0.0630689 | 0.5644342 | 86.948789 | 0.6131454 | 47.774087 | 16.735945 | 1.0324520 | 2.0621047 | 24.498551 | 7.0622567 | 98.318745 | 48.710537 | 0.6747912 | 58.900112 | 0.5758738 | 10.953921 |
| 12        | -0.486269 | 18623.190 | 0.2524634 | 44.413789 | 12.332376 | 2.5363318 | 1.1365637 | 0.2854343 | 2.8933507 | 4.3306441 | 8.1152377 | 9.6517510 | 12.446120 | 92.144988 | 25.185887 | 14.197725 | 86.284915 | 0.5332828 | 0.1032548 | 0.3089899 | 0.9358289 | 0.7597347 | 0.0357900 | 199.52972 | 0.4295239 | 46.095228 | 28.748747 | 0.1440186 | 7.8827366 | 18.567964 | 29.863812 | 9.6414160 | 76.805290 | 1.6964127 | 98.871299 | 0.1896098 | 11.561025 |
| 13        | -0.262148 | 50808.407 | 0.2626029 | 16.765209 | 9.2907019 | 2.8229727 | 2.6620298 | 0.3038538 | 3.5477343 | 1.0248625 | 2.0039942 | 12.704402 | 18.035346 | 91.387666 | 13.155174 | 19.364065 | 88.728962 | 0.3258323 | 0.5551846 | 0.1156528 | 0.9404530 | 0.7036374 | 0.1227053 | 112.29590 | 0.6754433 | 15.833477 | 11.611522 | 2.3267560 | 5.5475529 | 2.7503702 | 12.247924 | 12.230840 | 37.110322 | 1.3282923 | 142.10307 | 0.4520582 | 19.345381 |
| 14        | -0.218787 | 24809.246 | 0.2340791 | 17.808966 | 9.0594759 | 4.7386777 | 0.7134320 | 0.6385729 | 1.8135128 | 1.9133028 | 8.3340214 | 9.1700936 | 12.635080 | 91.047111 | 30.739300 | 14.045306 | 72.496493 | 0.9915159 | 0.5260855 | 0.1571770 | 0.9922728 | 0.0993921 | 0.2288328 | 176.10050 | 0.9277582 | 41.973817 | 40.886221 | 3.8471314 | 1.9154618 | 18.877098 | 6.0537812 | 52.744432 | 19.444956 | 1.5341300 | 66.826911 | 0.1498521 | 12.802440 |
| 15        | -0.201291 | 30257.529 | 0.0927589 | 48.586110 | 12.210501 | 2.8158924 | 1.9483848 | 0.7207387 | 1.7039725 | 2.6976016 | 8.8542309 | 9.6564870 | 14.592664 | 82.551498 | 43.033935 | 17.507872 | 85.794502 | 0.0850120 | 0.3465603 | 0.1634932 | 0.6374048 | 0.0681471 | 0.7003941 | 95.401181 | 0.3100850 | 11.274702 | 24.215312 | 2.9856762 | 7.0640232 | 5.2262545 | 28.962369 | 64.933656 | 51.159932 | 0.2179378 | 99.006206 | 0.2513735 | 4.7981906 |
| 16        | -0.215379 | 63404.671 | 0.0906472 | 31.113466 | 7.8370713 | 2.1935791 | 2.6824965 | 0.3416511 | 3.3423515 | 3.9308123 | 9.4276512 | 12.562236 | 12.408471 | 86.499601 | 61.127291 | 11.815775 | 82.620215 | 0.2070783 | 0.8623520 | 0.1973593 | 0.7000909 | 0.1608821 | 0.5653638 | 93.256152 | 0.1741060 | 5.6206504 | 11.536676 | 0.3432272 | 8.7461865 | 8.7061208 | 26.824351 | 52.632089 | 65.559237 | 2.6969423 | 12.423571 | 0.9213772 | 16.189062 |
| 17        | -0.486453 | 3566.3959 | 0.2473885 | 20.105816 | 12.599757 | 2.1451067 | 1.1834892 | 0.1606253 | 2.1338226 | 1.1895565 | 7.8452337 | 12.350337 | 5.4988647 | 77.856763 | 88.443560 | 17.183340 | 85.008691 | 0.4480438 | 0.8753708 | 0.2023456 | 0.9354244 | 0.2463339 | 0.9898504 | 157.49822 | 0.1367838 | 6.0409187 | 35.170552 | 0.2254529 | 3.3654175 | 4.5905862 | 27.521872 | 76.469037 | 65.544461 | 2.1390455 | 112.43511 | 0.7276474 | 17.084816 |
| 18        | -0.215379 | 31166.134 | 0.1276621 | 36.985640 | 7.9510520 | 3.5801010 | 2.1556190 | 0.8484570 | 1.9554943 | 4.9719055 | 3.4049450 | 7.6882431 | 18.942150 | 75.129182 | 36.385207 | 16.608581 | 87.590381 | 0.4644569 | 0.9093470 | 0.1067697 | 0.9706113 | 0.6549666 | 0.4552895 | 11.754302 | 0.3500005 | 15.928802 | 7.1360932 | 2.7311679 | 1.3762695 | 4.8134557 | 28.560461 | 96.934836 | 17.507448 | 2.8461307 | 89.504240 | 0.2575710 | 16.292643 |
| 19        | -0.516022 | 31983.253 | 0.1714929 | 29.223740 | 11.739988 | 3.7990912 | 1.6504482 | 0.9267783 | 2.7855197 | 4.1119995 | 4.0538643 | 9.6699210 | 16.908290 | 84.557237 | 46.688427 | 5.7621047 | 86.566238 | 0.9876870 | 0.5551083 | 0.1060758 | 0.7939727 | 0.9850980 | 0.0138236 | 73.074955 | 0.2476467 | 14.872908 | 19.951674 | 0.4854120 | 1.9419565 | 12.388767 | 20.639497 | 59.867959 | 59.014550 | 1.3477162 | 70.952052 | 0.0481024 | 11.846206 |
| 20        | -0.201291 | 67663.839 | 0.2706704 | 19.983499 | 10.462850 | 3.7987775 | 2.6595255 | 0.6373258 | 4.0388613 | 1.9763193 | 7.1509102 | 11.131235 | 6.7357542 | 73.419121 | 75.620057 | 9.7867752 | 89.561942 | 0.2325495 | 0.1715793 | 0.1290947 | 0.9586915 | 0.0729940 | 0.6183777 | 141.98301 | 0.9000498 | 41.442534 | 38.009221 | 1.6558564 | 5.0187019 | 25.183411 | 22.216205 | 75.685320 | 5.7018358 | 0.0547500 | 112.07361 | 0.7734399 | 24.630730 |
=========================================================================================================================================================================================================================================================================================================================================================================================================================================================================
Selected: robust_scaler
Held-out Protein: RMSE=0.0995, R²=0.9537

5. Preprocessing comparison — full spectra, inverted x-axis

from sklearn.pipeline import Pipeline

prep = Pipeline(best_pipeline.steps[:-1])
X_prep = np.asarray(prep.fit_transform(X))
chain = " → ".join(s for s, _ in best_pipeline.steps[:-1])

SpectroPrepPlotter.plot_preprocessing_comparison(
    wavelengths,
    X,
    {f"Optimal preprocessing ({chain})": X_prep},
    sample_indices=None,  # all 80 spectra
    title=f"Corn NIR (m5) — {component}: raw vs preprocessing",
    xlabel="Wavelength (nm)",
    alpha=0.3,
    invert_xaxis=True,
    figsize=(12, 8),
)
plt.show()

6. Predicted vs reference + optimisation diagnostics

SpectroPrepPlotter.plot_prediction_scatter(
    y[te], preds,
    title=f"Corn NIR (m5) — {component} (held-out demo)",
    xlabel=f"Reference {component}",
    ylabel=f"Predicted {component}",
)
SpectroPrepPlotter.plot_optimization_progress(opt)
SpectroPrepPlotter.plot_optimization_results(opt, top_n=5)
plt.show()

7. Paper figures (nested-CV, all constituents)

If run_corn_benchmark.py has been executed, display the saved prediction and preprocessing figures for every constituent (seed 0 of the ten-seed protocol).

from IPython.display import Image, display, Markdown

fig_dir = EXAMPLES / "corn_results" / "figures"
if fig_dir.exists():
    display(Markdown("### Raw spectra"))
    display(Image(filename=str(fig_dir / "spectra_raw_full.png")))
    for name in COMPONENTS:
        display(Markdown(f"### {name}"))
        pred = fig_dir / f"prediction_{name.lower()}.png"
        prep = fig_dir / f"spectra_prep_{name.lower()}.png"
        if pred.exists():
            display(Image(filename=str(pred)))
        if prep.exists():
            display(Image(filename=str(prep)))
else:
    print("Figures not found — run examples/run_corn_benchmark.py first.")

Raw spectra

Moisture

Oil

Protein

Starch