Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
236 lines (186 sloc) 7.9 KB
# ----------------------------------------------------------------------
# Numenta Platform for Intelligent Computing (NuPIC)
# Copyright (C) 2013, Numenta, Inc. Unless you have an agreement
# with Numenta, Inc., for a separate license for this software code, the
# following terms and conditions apply:
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero Public License version 3 as
# published by the Free Software Foundation.
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# See the GNU Affero Public License for more details.
# You should have received a copy of the GNU Affero Public License
# along with this program. If not, see
# ----------------------------------------------------------------------
from nupic.encoders.base import Encoder
from nupic.encoders import (ScalarEncoder,
import capnp
except ImportError:
capnp = None
if capnp:
from nupic.encoders.multi_capnp import MultiEncoderProto
# Map class to Cap'n Proto schema union attribute
ScalarEncoder: "scalarEncoder",
AdaptiveScalarEncoder: "adaptiveScalarEncoder",
DateEncoder: "dateEncoder",
LogEncoder: "logEncoder",
CategoryEncoder: "categoryEncoder",
CoordinateEncoder: "coordinateEncoder",
SDRCategoryEncoder: "sdrCategoryEncoder",
DeltaEncoder: "deltaEncoder",
PassThroughEncoder: "passThroughEncoder",
SparsePassThroughEncoder: "sparsePassThroughEncoder",
GeospatialCoordinateEncoder: "geospatialCoordinateEncoder",
ScalarSpaceEncoder: "scalarSpaceEncoder",
RandomDistributedScalarEncoder: "randomDistributedScalarEncoder"
# Invert for fast lookup in
_ATTR_CLASS_MAP = {value: key for key, value in _CLASS_ATTR_MAP.items()}
class MultiEncoder(Encoder):
A MultiEncoder encodes a dictionary or object with multiple components. A
MultiEncoder contains a number of sub-encoders, each of which encodes a
separate component.
:param encoderDefinitions: a dict of dicts, mapping field names to the field
params dict. Sent directly to :meth:`.addMultipleEncoders`.
def __init__(self, encoderDefinitions=None):
self.width = 0
self.encoders = []
self.description = [] = ''
self._flattenedEncoderList = None
self._flattenedFieldTypeList = None
if encoderDefinitions is not None:
def setFieldStats(self, fieldName, fieldStatistics ):
for (name, encoder, offset) in self.encoders:
encoder.setFieldStats(name, fieldStatistics)
def addEncoder(self, name, encoder):
Adds one encoder.
:param name: (string) name of encoder, should be unique
:param encoder: (:class:`.Encoder`) the encoder to add
self.encoders.append((name, encoder, self.width))
for d in encoder.getDescription():
self.description.append((d[0], d[1] + self.width))
self.width += encoder.getWidth()
def encodeIntoArray(self, obj, output):
for name, encoder, offset in self.encoders:
encoder.encodeIntoArray(self._getInputValue(obj, name), output[offset:])
def getDescription(self):
return self.description
def getWidth(self):
"""Represents the sum of the widths of each fields encoding."""
return self.width
def setLearning(self,learningEnabled):
encoders = self.getEncoderList()
for encoder in encoders:
def encodeField(self, fieldName, value):
for name, encoder, offset in self.encoders:
if name == fieldName:
return encoder.encode(value)
def encodeEachField(self, inputRecord):
encodings = []
for name, encoder, offset in self.encoders:
encodings.append(encoder.encode(getattr(inputRecord, name)))
return encodings
def addMultipleEncoders(self, fieldEncodings):
:param fieldEncodings: dict of dicts, mapping field names to the field
params dict.
Each field params dict has the following keys:
1. ``fieldname``: data field name
2. ``type`` an encoder type
3. All other keys are encoder parameters
For example,
.. code-block:: python
'dateTime': dict(fieldname='dateTime', type='DateEncoder',
'attendeeCount': dict(fieldname='attendeeCount', type='ScalarEncoder',
name='attendeeCount', minval=0, maxval=250,
clipInput=True, w=5, resolution=10),
'consumption': dict(fieldname='consumption',type='ScalarEncoder',
name='consumption', minval=0,maxval=110,
clipInput=True, w=5, resolution=5),
would yield a vector with a part encoded by the :class:`.DateEncoder`, and
to parts seperately taken care of by the :class:`.ScalarEncoder` with the
specified parameters. The three seperate encodings are then merged together
to the final vector, in such a way that they are always at the same location
within the vector.
# Sort the encoders so that they end up in a controlled order
encoderList = sorted(fieldEncodings.items())
for key, fieldParams in encoderList:
if ':' not in key and fieldParams is not None:
fieldParams = fieldParams.copy()
fieldName = fieldParams.pop('fieldname')
encoderName = fieldParams.pop('type')
self.addEncoder(fieldName, eval(encoderName)(**fieldParams))
except TypeError, e:
print ("#### Error in constructing %s encoder. Possibly missing "
"some required constructor parameters. Parameters "
"that were provided are: %s" % (encoderName, fieldParams))
def getSchema(cls):
return MultiEncoderProto
def read(cls, proto):
encoder = object.__new__(cls)
encoder._flattenedEncoderList = None
encoder._flattenedFieldTypeList = None
encoder.encoders = [None] * len(proto.encoders)
encoder.width = 0
for index, encoderProto in enumerate(proto.encoders):
# Identify which attr is set in union
encoderType = encoderProto.which()
encoderDetails = getattr(encoderProto, encoderType)
encoder.encoders[index] = (,
# Call where class is determined by _ATTR_CLASS_MAP
encoder.width += encoder.encoders[index][1].getWidth()
# Derive description from encoder list
encoder.description = [(enc[1].name, int(enc[2]))
for enc in encoder.encoders] =
return encoder
def write(self, proto):
proto.init("encoders", len(self.encoders))
for index, (name, encoder, offset) in enumerate(self.encoders):
encoderProto = proto.encoders[index]
encoderType = _CLASS_ATTR_MAP.get(encoder.__class__)
encoderDetails = getattr(encoderProto, encoderType)
encoder.write(encoderDetails) = name
encoderProto.offset = offset =
You can’t perform that action at this time.