Package rdkit :: Package ML :: Package Data :: Module Quantize

Module Quantize

Automatic search for quantization bounds

This uses the expected informational gain to determine where quantization bounds should
lie.

**Notes**:

  - bounds are less than, so if the bounds are [1.,2.],
    [0.9,1.,1.1,2.,2.2] -> [0,1,1,2,2]

Functions

[hide private]

feq(v1, v2, tol=_float_tol)
floating point equality with a tolerance factor

source code

FindVarQuantBound(vals, results, nPossibleRes)
Uses FindVarMultQuantBounds, only here for historic reasons...

source code

_GenVarTable(vals, cuts, starts, results, nPossibleRes)
Primarily intended for internal use

source code

_PyRecurseOnBounds(vals, cuts, which, starts, results, nPossibleRes, varTable=None)
Primarily intended for internal use

source code

_NewPyRecurseOnBounds(vals, cuts, which, starts, results, nPossibleRes, varTable=None)
Primarily intended for internal use

source code

_NewPyFindStartPoints(sortVals, sortResults, nData)

source code

FindVarMultQuantBounds(vals, nBounds, results, nPossibleRes)
finds multiple quantization bounds for a single variable

source code

_RecurseOnBounds(vals, cuts, which, starts, results, nPossibleRes, varTable=None)
Primarily intended for internal use

source code

_FindStartPoints(sortVals, sortResults, nData)

source code

Variables

[hide private]

hascQuantize = 1

_float_tol = 1e-8

Imports: numpy, entropy, zip, map, range, cQuantize

Function Details

[hide private]

feq(v1, v2, tol=_float_tol)

source code

floating point equality with a tolerance factor

**Arguments**

  - v1: a float

  - v2: a float

  - tol: the tolerance for comparison

**Returns**

  0 or 1

FindVarQuantBound(vals, results, nPossibleRes)

source code

Uses FindVarMultQuantBounds, only here for historic reasons

_GenVarTable(vals, cuts, starts, results, nPossibleRes)

source code

Primarily intended for internal use

constructs a variable table for the data passed in
The table for a given variable records the number of times each possible value
 of that variable appears for each possible result of the function.

**Arguments**

  - vals: a 1D Numeric array with the values of the variables

  - cuts: a list with the indices of the quantization bounds
    (indices are into _starts_ )

  - starts: a list of potential starting points for quantization bounds

  - results: a 1D Numeric array of integer result codes

  - nPossibleRes: an integer with the number of possible result codes

**Returns**

  the varTable, a 2D Numeric array which is nVarValues x nPossibleRes

**Notes**

  - _vals_ should be sorted!

_PyRecurseOnBounds(vals, cuts, which, starts, results, nPossibleRes, varTable=None)

source code

Primarily intended for internal use

Recursively finds the best quantization boundaries

**Arguments**

  - vals: a 1D Numeric array with the values of the variables,
    this should be sorted

  - cuts: a list with the indices of the quantization bounds
    (indices are into _starts_ )

  - which: an integer indicating which bound is being adjusted here
    (and index into _cuts_ )

  - starts: a list of potential starting points for quantization bounds

  - results: a 1D Numeric array of integer result codes

  - nPossibleRes: an integer with the number of possible result codes

**Returns**

  - a 2-tuple containing:

    1) the best information gain found so far

    2) a list of the quantization bound indices ( _cuts_ for the best case)

**Notes**

 - this is not even remotely efficient, which is why a C replacement
   was written

_NewPyRecurseOnBounds(vals, cuts, which, starts, results, nPossibleRes, varTable=None)

source code

Primarily intended for internal use

Recursively finds the best quantization boundaries

**Arguments**

  - vals: a 1D Numeric array with the values of the variables,
    this should be sorted

  - cuts: a list with the indices of the quantization bounds
    (indices are into _starts_ )

  - which: an integer indicating which bound is being adjusted here
    (and index into _cuts_ )

  - starts: a list of potential starting points for quantization bounds

  - results: a 1D Numeric array of integer result codes

  - nPossibleRes: an integer with the number of possible result codes

**Returns**

  - a 2-tuple containing:

    1) the best information gain found so far

    2) a list of the quantization bound indices ( _cuts_ for the best case)

**Notes**

 - this is not even remotely efficient, which is why a C replacement
   was written

FindVarMultQuantBounds(vals, nBounds, results, nPossibleRes)

source code

finds multiple quantization bounds for a single variable

**Arguments**

  - vals: sequence of variable values (assumed to be floats)

  - nBounds: the number of quantization bounds to find

  - results: a list of result codes (should be integers)

  - nPossibleRes: an integer with the number of possible values of the
    result variable

**Returns**

  - a 2-tuple containing:

    1) a list of the quantization bounds (floats)

    2) the information gain associated with this quantization

_RecurseOnBounds(vals, cuts, which, starts, results, nPossibleRes, varTable=None)

source code

Primarily intended for internal use

Recursively finds the best quantization boundaries

**Arguments**

  - vals: a 1D Numeric array with the values of the variables,
    this should be sorted

  - cuts: a list with the indices of the quantization bounds
    (indices are into _starts_ )

  - which: an integer indicating which bound is being adjusted here
    (and index into _cuts_ )

  - starts: a list of potential starting points for quantization bounds

  - results: a 1D Numeric array of integer result codes

  - nPossibleRes: an integer with the number of possible result codes

**Returns**

  - a 2-tuple containing:

    1) the best information gain found so far

    2) a list of the quantization bound indices ( _cuts_ for the best case)

**Notes**

 - this is not even remotely efficient, which is why a C replacement
   was written