Overview | Background: NIR spectroscopy and applications | Application workflows for ProxiMate and ProxiScout | Scope of proximetricsR

Introduction | Correlation vs. covariance | Partial Least Squares Regression algorithms | Standard PLSR algorithm | Modified PLSR algorithm | NIRWise PLUS-compatible PLSR implementation | Derived matrices and diagnostics | PLS predictions | Predictions using scores | Predictions using regression coefficients | Extended PLSR: the XLS algorithm | References

Introduction | Build an application | Read the calibration data and prepare it | Merge multiple datasets | Resample to constant resolution | Activate/deactivate rows for modeling | Get the names of the response variables | Calibrating models using calibrate_models() | formulas: defining what is to be modeled | metadata_list: specify the property metadata | preprocess_recipes: create pre-processing recipes | method: how to fit the spectral models | control: Setting up the calibration parameters | Calibrate the spectral models | Overview of the best models found | Checking all the models tested | Predicting the properties in unseen samples | Writting down an application file | Other functionality | Calibrate single models with calibrate | Write model-related files (tsv, cal, prj and rtf)

Introduction | Structure of the ProxiMate predictive applications | Calibration data file (.tsv) | Local data file (.tsv) | Calibration model files (.cal) | Project files (.prj) | Report files (.rtf) | Application metadata file (.nad) | Application file (.nax) | References

Introduction | Setup | Workflow overview | Prepare spectral data | Define preprocessing recipe | Build calibration model | Example 1: Build a single model | Serialize the model for deployment | Example 2: Build multiple models at once testing different pre-processings | Serialize the multiple models for deployment | Export for ProxiScout | Device-specific considerations | References

Introduction | ProxiScout predictive application package | Overview | Deployment workflow | operations.json | Purpose | File structure | Operation Object | Example | Execution model | proximetricsR function reference | Spectra scale (ID: 37) | Get absorbance (ID: 29) | Average readings (ID: 7) | SNV (ID: 2) | Detrending (ID: 3) | Savitzky-Golay smoothing and differentiation (ID: 83) | Variable selection (ID: 17) | model_info.json | Contents

Summary | Calibrate an application and write a nax | Read ProxiMate application files (.nax) | Recalibrate application | Just re-fit the models | Recalibrate based on new data

Overview | Key concepts | Setup | Preprocessing constructors | Resampling: prep_resample() | Smoothing: prep_smooth() | Standard Normal Variate: prep_snv() | Derivatives: prep_derivative() | Detrending: prep_detrend() | Reflectance/Absorbance conversion: prep_transform() | Wavelength trimming: prep_wav_trim() | Building preprocessing recipes | Device compatibility | Building recipes | Applying recipes with process() | Practical examples | Example 1: ProxiMate workflow | Example 2: ProxiScout workflow with detrending | Example 3: Minimal preprocessing | Example 4: Wavelength band selection | Best practices | Order matters | Device-aware development | Reproducibility | Parameter tuning | Summary

Introduction | Noise removal | Moving average | Savitzky-Golay filtering | Derivatives | Scatter and baseline corrections | Standard Normal Variate (SNV) | Multiplicative Scatter Correction (MSC) | SNV-Detrend | Baseline removal | Centering and scaling | Resampling | Other transformations | Continuum removal | References

Preamble | Introduction | Citing the package | Further reading | References

Introduction | $k$-means sampling (naes) | Kennard-Stone sampling (kenStone) | DUPLEX (duplex) | SELECT algorithm (shenkWest) | Puchwein algorithm (puchwein) | Honigs sampling (honigs) | References

Introduction | Methods available in the resemble package | Selecting the number of components | Variance-based selection | Individual variance threshold (ncomp_by_var()) | Cumulative variance threshold (ncomp_by_cumvar()) | Optimal component selection (ncomp_by_opc()) | Fixed number of components | Dimentionality reduction methods in ortho_projection() | Available methods | Example: PCA projection | Example: PLS projection | Example: Optimal component selection | Projecting new data | Multivariate side information | References

Introduction | Citing the package | Dataset used across the vignettes | Spectral preprocessing | References

Introduction | Main notation | Available dissimilarity construtors | Dissimilarity in orthogonal projection spaces | PCA-based dissimilarity | PLS-based dissimilarity | Correlation dissimilarity | Conventional correlation dissimilarity | Moving window correlation dissimilarity | Parallel execution for correlation-based dissimilarity (advanced users) | Euclidean dissimilarity | Cosine dissimilarity | Evaluating dissimilarity methods | References

Searching for neighbors | Neighbor selection methods | Fixed-$k$ selection | Dissimilarity-threshold selection | Using a specific number of neighbors | Using a dissimilarity threshold | Spiking the neighborhoods

Introduction | Fitting methods | Example | Data preparation | Fitting a PLS model | Fitting a GPR model | Prediction | Validation statistics | References

Memory-based learning | Fitting methods and constructors | The mbl_control() function and validation types | Model fitting and prediction | Additional examples | Using Yu argument | Supported parallel processing | References

Introduction | Glossary and conventions | The gesearch algorithm: How it works | Evolutionary search overview | Weakness score computation | Step 1: Evaluate each individual | Step 2: Aggregate to gene-level weakness | Interpretation | Weakness functions | Gene silencing | Final model | Key parameters | Practical considerations | Example workflow | The dataset, the target set, and its practical constraints | A simple linear model for predicting in the target set (baseline) | The model built with the samples found by gesearch | Further examples of gesearch | No response values available for the target set | No response values in the target set but with information about the response range | Supported parallel processing | References

Introduction | How liblex works | Build phase | Prediction phase | Data preparation | Core components | Neighbor selection | Dissimilarity methods | Fitting method | Control parameters | Building a library of models | Using fixed neighborhood size(s) | Using dissimilarity thresholds | Visualizing neighborhood centroids | Visualizing the regression models | Choosing anchor samples | Using k-means sampling | Building the library with selected anchors | Prediction | Prediction output | Weighting options | Enforcing specific experts | Validation and performance | Evaluation metrics | Uncertainty as a quality filter | Visualization | Comparison with classical MBL | Parallel processing | Summary | References