Generating Bounded Model Inputs

After initial normalization, a feature can still be unbounded or strongly skewed. The map classes in this module define the transformations from the normalized features to the final inputs used by the ML model or mapped evaluator. Each map records its raw input indices and parameters, evaluates one transformed feature, and propagates its derivative back to every raw input it uses.

FeatureList stores these maps in the order expected from the model. fill_vals_ applies the forward maps, fill_derivs_ applies their adjoints, and YAML serialization preserves map types and parameters. Examples include the rational map \(y=\gamma x/(1+\gamma x)\) and the SLTMap used for the bounded CIDER26XC kinetic-energy indicator. Model loading supplies the exact list associated with a packaged functional.

class ciderpress.dft.transform_data.FeatureList(feat_list)

class ciderpress.dft.transform_data.EMap(i, scale=1.0, center=0.0, bounds=None)

class ciderpress.dft.transform_data.FeatureNormalizer

class ciderpress.dft.transform_data.LMap(i, bounds=None)

class ciderpress.dft.transform_data.OmegaMap(i_n, i_s, i_alpha, c, B, C, bounds=None)

class ciderpress.dft.transform_data.SLBMap(i, j, k)

Map \(\beta=(\tau-\tau_{\mathrm W})/(\tau+\tau_0)\).

class ciderpress.dft.transform_data.SLDMap(i, j, k)

class ciderpress.dft.transform_data.SLNMap(i, gamma)

class ciderpress.dft.transform_data.SLR2Map(i, j)

class ciderpress.dft.transform_data.SLR3Map(i, j)

class ciderpress.dft.transform_data.SLRMap(i, j, k)

class ciderpress.dft.transform_data.SLTMap(i, j)

Map \(\beta=(\tau-\tau_0)/(\tau+\tau_0)\).

class ciderpress.dft.transform_data.SLTWMap(i, j)

Map \(\beta=(\tau_{\mathrm W}-\tau_0)/(\tau_{\mathrm W}+\tau_0)\).

class ciderpress.dft.transform_data.SLXMap(i, j, gamma)

Map the squared reduced gradient to a bounded feature.

The map is \(\gamma s^2/(1+\gamma s^2)\), where \(s^2=\sigma/(C n^{8/3})\) is the squared reduced gradient.

class ciderpress.dft.transform_data.SignedUMap(i, gamma)

class ciderpress.dft.transform_data.TMap(i, j, bounds=None)

class ciderpress.dft.transform_data.UMap(i, gamma, bounds=None)

class ciderpress.dft.transform_data.V2Map(i, j)

class ciderpress.dft.transform_data.V3Map(i, j, gamma)

class ciderpress.dft.transform_data.V4Map(i, j, gamma)

class ciderpress.dft.transform_data.VMap(i, gamma, scale=1.0, center=0.0, bounds=None)

class ciderpress.dft.transform_data.VZMap(i, gamma, scale=1.0, center=0.0, bounds=None)

class ciderpress.dft.transform_data.WMap(i, j, k, gammai, gammaj)

class ciderpress.dft.transform_data.XMap(i, j, k, gammai, gammaj)

class ciderpress.dft.transform_data.YMap(i, j, k, l, gammai, gammaj, gammak)

class ciderpress.dft.transform_data.ZMap(i, gamma, scale=1.0, center=0.0, bounds=None)

ciderpress.dft.transform_data.cls

alias of OmegaMap