Skip to content

Decoding API Reference

Bayesian encoding and decoding

BayesianDecoder

Bases: object

Bayesian decoder for neural population activity.

This class provides a scikit-learn-like API for Bayesian decoding using tuning curves. Supports 1D and 2D decoding, and can be extended for more complex models.

Parameters:

Name Type Description Default
tuningcurve TuningCurve1D or TuningCurve2D

Tuning curve to use for decoding.

 None
Source code in nelpy/decoding.py 
 863
 864
 865
 866
 867
 868
 869
 870
 871
 872
 873
 874
 875
 876
 877
 878
 879
 880
 881
 882
 883
 884
 885
 886
 887
 888
 889
 890
 891
 892
 893
 894
 895
 896
 897
 898
 899
 900
 901
 902
 903
 904
 905
 906
 907
 908
 909
 910
 911
 912
 913
 914
 915
 916
 917
 918
 919
 920
 921
 922
 923
 924
 925
 926
 927
 928
 929
 930
 931
 932
 933
 934
 935
 936
 937
 938
 939
 940
 941
 942
 943
 944
 945
 946
 947
 948
 949
 950
 951
 952
 953
 954
 955
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
class BayesianDecoder(object):
    """
    Bayesian decoder for neural population activity.

    This class provides a scikit-learn-like API for Bayesian decoding using tuning curves.
    Supports 1D and 2D decoding, and can be extended for more complex models.

    Parameters
    ----------
    tuningcurve : TuningCurve1D or TuningCurve2D, optional
        Tuning curve to use for decoding.
    """

    def __init__(self, tuningcurve=None):
        """
        Initialize the BayesianDecoder.

        Parameters
        ----------
        tuningcurve : TuningCurve1D or TuningCurve2D, optional
            Tuning curve to use for decoding.
        """
        self.tuningcurve = tuningcurve

    def fit(self, X, y=None):
        """
        Fit the decoder to data X. (Stores the tuning curve if provided.)

        Parameters
        ----------
        X : array-like or TuningCurve1D/2D
            Training data or tuning curve.
        y : Ignored
        """
        # If X is a tuning curve, store it
        self.tuningcurve = X
        return self

    def predict_proba(self, X, **kwargs):
        """
        Predict posterior probabilities for data X.

        Parameters
        ----------
        X : array-like or BinnedEventArray
            Data to decode.
        Returns
        -------
        posterior : np.ndarray
            Posterior probability matrix.
        """
        if self.tuningcurve is None:
            raise ValueError(
                "No tuning curve set. Call fit() or provide tuningcurve in constructor."
            )
        # Use decode1D or decode2D depending on tuning curve
        if hasattr(self.tuningcurve, "ratemap") and hasattr(self.tuningcurve, "bins"):
            # TuningCurve1D or TuningCurve2D object
            ratemap = self.tuningcurve.ratemap
        else:
            ratemap = self.tuningcurve
        # Try to infer 1D vs 2D
        if ratemap.ndim == 2:
            posterior, _, _, _ = decode1D(X, ratemap, **kwargs)
        elif ratemap.ndim == 3:
            posterior, _, _, _ = decode2D(X, ratemap, **kwargs)
        else:
            raise ValueError("Tuning curve must be 2D or 3D array.")
        return posterior

    def predict(self, X, **kwargs):
        """
        Predict external variable from data X (returns mode path).

        Parameters
        ----------
        X : array-like or BinnedEventArray
            Data to decode.
        Returns
        -------
        mode_pth : np.ndarray
            Most likely position at each time bin.
        """
        if self.tuningcurve is None:
            raise ValueError(
                "No tuning curve set. Call fit() or provide tuningcurve in constructor."
            )
        if hasattr(self.tuningcurve, "ratemap") and hasattr(self.tuningcurve, "bins"):
            ratemap = self.tuningcurve.ratemap
        else:
            ratemap = self.tuningcurve
        if ratemap.ndim == 2:
            _, _, mode_pth, _ = decode1D(X, ratemap, **kwargs)
        elif ratemap.ndim == 3:
            _, _, mode_pth, _ = decode2D(X, ratemap, **kwargs)
        else:
            raise ValueError("Tuning curve must be 2D or 3D array.")
        return mode_pth

    def predict_asa(self, X, **kwargs):
        """
        Predict analog signal array (mean path) from data X.

        Parameters
        ----------
        X : array-like or BinnedEventArray
            Data to decode.
        Returns
        -------
        asa : AnalogSignalArray or np.ndarray
            Mean path as AnalogSignalArray if possible, else array.
        """
        if self.tuningcurve is None:
            raise ValueError(
                "No tuning curve set. Call fit() or provide tuningcurve in constructor."
            )
        if hasattr(self.tuningcurve, "ratemap") and hasattr(self.tuningcurve, "bins"):
            ratemap = self.tuningcurve.ratemap
        else:
            ratemap = self.tuningcurve
        if ratemap.ndim == 2:
            _, _, _, mean_pth = decode1D(X, ratemap, **kwargs)
        elif ratemap.ndim == 3:
            _, _, _, mean_pth = decode2D(X, ratemap, **kwargs)
        else:
            raise ValueError("Tuning curve must be 2D or 3D array.")
        # Try to return as AnalogSignalArray if possible, with timestamps if available
        try:
            from .core import AnalogSignalArray

            abscissa_vals = None
            # Try to get bin centers from X if possible
            if hasattr(X, "bin_centers"):
                abscissa_vals = X.bin_centers
            elif hasattr(X, "abscissa_vals"):
                abscissa_vals = X.abscissa_vals
            return AnalogSignalArray(data=mean_pth, abscissa_vals=abscissa_vals)
        except Exception:
            return mean_pth

    def __repr__(self):
        s = "<BayesianDecoder: "
        if self.tuningcurve is not None:
            s += f"tuningcurve shape={getattr(self.tuningcurve, 'ratemap', self.tuningcurve).shape}"
        else:
            s += "no tuningcurve"
        s += ">"
        return s

fit(X, y=None)

Fit the decoder to data X. (Stores the tuning curve if provided.)

Parameters:

Name Type Description Default
X array-like or TuningCurve1D/2D

Training data or tuning curve.

 required
y Ignored
None
Source code in nelpy/decoding.py 
887
888
889
890
891
892
893
894
895
896
897
898
899
def fit(self, X, y=None):
    """
    Fit the decoder to data X. (Stores the tuning curve if provided.)

    Parameters
    ----------
    X : array-like or TuningCurve1D/2D
        Training data or tuning curve.
    y : Ignored
    """
    # If X is a tuning curve, store it
    self.tuningcurve = X
    return self

predict(X, **kwargs)

Predict external variable from data X (returns mode path).

Parameters:

Name Type Description Default
X array - like or BinnedEventArray

Data to decode.

 required

Returns:

Name Type Description
mode_pth ndarray

Most likely position at each time bin.
Source code in nelpy/decoding.py 
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
def predict(self, X, **kwargs):
    """
    Predict external variable from data X (returns mode path).

    Parameters
    ----------
    X : array-like or BinnedEventArray
        Data to decode.
    Returns
    -------
    mode_pth : np.ndarray
        Most likely position at each time bin.
    """
    if self.tuningcurve is None:
        raise ValueError(
            "No tuning curve set. Call fit() or provide tuningcurve in constructor."
        )
    if hasattr(self.tuningcurve, "ratemap") and hasattr(self.tuningcurve, "bins"):
        ratemap = self.tuningcurve.ratemap
    else:
        ratemap = self.tuningcurve
    if ratemap.ndim == 2:
        _, _, mode_pth, _ = decode1D(X, ratemap, **kwargs)
    elif ratemap.ndim == 3:
        _, _, mode_pth, _ = decode2D(X, ratemap, **kwargs)
    else:
        raise ValueError("Tuning curve must be 2D or 3D array.")
    return mode_pth

predict_asa(X, **kwargs)

Predict analog signal array (mean path) from data X.

Parameters:

Name Type Description Default
X array - like or BinnedEventArray

Data to decode.

 required

Returns:

Name Type Description
asa AnalogSignalArray or ndarray

Mean path as AnalogSignalArray if possible, else array.
Source code in nelpy/decoding.py 
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
def predict_asa(self, X, **kwargs):
    """
    Predict analog signal array (mean path) from data X.

    Parameters
    ----------
    X : array-like or BinnedEventArray
        Data to decode.
    Returns
    -------
    asa : AnalogSignalArray or np.ndarray
        Mean path as AnalogSignalArray if possible, else array.
    """
    if self.tuningcurve is None:
        raise ValueError(
            "No tuning curve set. Call fit() or provide tuningcurve in constructor."
        )
    if hasattr(self.tuningcurve, "ratemap") and hasattr(self.tuningcurve, "bins"):
        ratemap = self.tuningcurve.ratemap
    else:
        ratemap = self.tuningcurve
    if ratemap.ndim == 2:
        _, _, _, mean_pth = decode1D(X, ratemap, **kwargs)
    elif ratemap.ndim == 3:
        _, _, _, mean_pth = decode2D(X, ratemap, **kwargs)
    else:
        raise ValueError("Tuning curve must be 2D or 3D array.")
    # Try to return as AnalogSignalArray if possible, with timestamps if available
    try:
        from .core import AnalogSignalArray

        abscissa_vals = None
        # Try to get bin centers from X if possible
        if hasattr(X, "bin_centers"):
            abscissa_vals = X.bin_centers
        elif hasattr(X, "abscissa_vals"):
            abscissa_vals = X.abscissa_vals
        return AnalogSignalArray(data=mean_pth, abscissa_vals=abscissa_vals)
    except Exception:
        return mean_pth

predict_proba(X, **kwargs)

Predict posterior probabilities for data X.

Parameters:

Name Type Description Default
X array - like or BinnedEventArray

Data to decode.

 required

Returns:

Name Type Description
posterior ndarray

Posterior probability matrix.
Source code in nelpy/decoding.py 
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
def predict_proba(self, X, **kwargs):
    """
    Predict posterior probabilities for data X.

    Parameters
    ----------
    X : array-like or BinnedEventArray
        Data to decode.
    Returns
    -------
    posterior : np.ndarray
        Posterior probability matrix.
    """
    if self.tuningcurve is None:
        raise ValueError(
            "No tuning curve set. Call fit() or provide tuningcurve in constructor."
        )
    # Use decode1D or decode2D depending on tuning curve
    if hasattr(self.tuningcurve, "ratemap") and hasattr(self.tuningcurve, "bins"):
        # TuningCurve1D or TuningCurve2D object
        ratemap = self.tuningcurve.ratemap
    else:
        ratemap = self.tuningcurve
    # Try to infer 1D vs 2D
    if ratemap.ndim == 2:
        posterior, _, _, _ = decode1D(X, ratemap, **kwargs)
    elif ratemap.ndim == 3:
        posterior, _, _, _ = decode2D(X, ratemap, **kwargs)
    else:
        raise ValueError("Tuning curve must be 2D or 3D array.")
    return posterior

Cumhist

Bases: ndarray

Cumulative histogram with interpolation support.

Parameters:

Name Type Description Default
cumhist ndarray

Cumulative histogram values.

 required
bincenters ndarray

Bin centers corresponding to the cumulative histogram.

 required
Source code in nelpy/decoding.py 
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
class Cumhist(np.ndarray):
    """
    Cumulative histogram with interpolation support.

    Parameters
    ----------
    cumhist : np.ndarray
        Cumulative histogram values.
    bincenters : np.ndarray
        Bin centers corresponding to the cumulative histogram.
    """

    def __new__(cls, cumhist, bincenters):
        obj = np.asarray(cumhist).view(cls)
        obj._bincenters = bincenters
        return obj

    def __call__(self, *val):
        f = interpolate.interp1d(
            x=self, y=self._bincenters, kind="linear", fill_value=np.nan
        )
        try:
            vals = f(*val).item()
        except AttributeError:
            vals = f(*val)

        return vals

ItemGetter_iloc

Bases: object

.iloc is primarily integer position based (from 0 to length-1 of the axis).

Allows integer-based selection of intervals and series in event arrays. Raises IndexError if a requested indexer is out-of-bounds, except slice indexers which allow out-of-bounds indexing (conforms with python/numpy slice semantics).

Allowed inputs are: - An integer e.g. 5 - A list or array of integers [4, 3, 0] - A slice object with ints 1:7

Parameters:

Name Type Description Default
obj object

The parent object to slice.

 required
Source code in nelpy/decoding.py 
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
class ItemGetter_iloc(object):
    """
    .iloc is primarily integer position based (from 0 to length-1 of the axis).

    Allows integer-based selection of intervals and series in event arrays.
    Raises IndexError if a requested indexer is out-of-bounds, except slice indexers which allow out-of-bounds indexing (conforms with python/numpy slice semantics).

    Allowed inputs are:
        - An integer e.g. 5
        - A list or array of integers [4, 3, 0]
        - A slice object with ints 1:7

    Parameters
    ----------
    obj : object
        The parent object to slice.
    """

    def __init__(self, obj):
        self.obj = obj

    def __getitem__(self, idx):
        """intervals, series"""
        intervalslice, seriesslice = self.obj._slicer[idx]
        out = copy.copy(self.obj)
        if isinstance(seriesslice, int):
            seriesslice = [seriesslice]
        out._data = out._data[seriesslice]
        singleseries = len(out._data) == 1
        if singleseries:
            out._data = np.array(out._data[0], ndmin=2)
        out._series_ids = list(
            np.atleast_1d(np.atleast_1d(out._series_ids)[seriesslice])
        )
        out._series_labels = list(
            np.atleast_1d(np.atleast_1d(out._series_labels)[seriesslice])
        )
        # TODO: update tags
        if isinstance(intervalslice, slice):
            if (
                intervalslice.start is None
                and intervalslice.stop is None
                and intervalslice.step is None
            ):
                out.loc = ItemGetter_loc(out)
                out.iloc = ItemGetter_iloc(out)
                return out
        out = out._intervalslicer(intervalslice)
        out.loc = ItemGetter_loc(out)
        out.iloc = ItemGetter_iloc(out)
        return out

ItemGetter_loc

Bases: object

.loc is primarily label based (that is, series_id based).

Allows label-based selection of intervals and series in event arrays. Raises KeyError when the items are not found.

Allowed inputs are: - A single label, e.g. 5 or 'a' (interpreted as a label, not a position) - A list or array of labels ['a', 'b', 'c'] - A slice object with labels 'a':'f' (both start and stop are included)

Parameters:

Name Type Description Default
obj object

The parent object to slice.

 required
Source code in nelpy/decoding.py 
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
class ItemGetter_loc(object):
    """
    .loc is primarily label based (that is, series_id based).

    Allows label-based selection of intervals and series in event arrays.
    Raises KeyError when the items are not found.

    Allowed inputs are:
        - A single label, e.g. 5 or 'a' (interpreted as a label, not a position)
        - A list or array of labels ['a', 'b', 'c']
        - A slice object with labels 'a':'f' (both start and stop are included)

    Parameters
    ----------
    obj : object
        The parent object to slice.
    """

    def __init__(self, obj):
        self.obj = obj

    def __getitem__(self, idx):
        """intervals, series"""
        intervalslice, seriesslice = self.obj._slicer[idx]

        # first convert series slice into list
        if isinstance(seriesslice, slice):
            start = seriesslice.start
            stop = seriesslice.stop
            istep = seriesslice.step
            try:
                if start is None:
                    istart = 0
                else:
                    istart = self.obj._series_ids.index(start)
            except ValueError:
                raise KeyError(
                    "series_id {} could not be found in BaseEventArray!".format(start)
                )
            try:
                if stop is None:
                    istop = self.obj.n_series
                else:
                    istop = self.obj._series_ids.index(stop) + 1
            except ValueError:
                raise KeyError(
                    "series_id {} could not be found in BaseEventArray!".format(stop)
                )
            if istep is None:
                istep = 1
            if istep < 0:
                istop -= 1
                istart -= 1
                istart, istop = istop, istart
            series_idx_list = list(range(istart, istop, istep))
        else:
            series_idx_list = []
            seriesslice = np.atleast_1d(seriesslice)
            for series in seriesslice:
                try:
                    uidx = self.obj.series_ids.index(series)
                except ValueError:
                    raise KeyError(
                        "series_id {} could not be found in BaseEventArray!".format(
                            series
                        )
                    )
                else:
                    series_idx_list.append(uidx)

        if not isinstance(series_idx_list, list):
            series_idx_list = list(series_idx_list)
        out = copy.copy(self.obj)
        try:
            out._data = out._data[series_idx_list]
            singleseries = len(out._data) == 1
        except AttributeError:
            out._data = out._data[series_idx_list]
            singleseries = len(out._data) == 1

        if singleseries:
            out._data = np.array(out._data[0], ndmin=2)
        out._series_ids = list(
            np.atleast_1d(np.atleast_1d(out._series_ids)[series_idx_list])
        )
        out._series_labels = list(
            np.atleast_1d(np.atleast_1d(out._series_labels)[series_idx_list])
        )
        # TODO: update tags
        if isinstance(intervalslice, slice):
            if (
                intervalslice.start is None
                and intervalslice.stop is None
                and intervalslice.step is None
            ):
                out.loc = ItemGetter_loc(out)
                out.iloc = ItemGetter_iloc(out)
                return out
        out = out._intervalslicer(intervalslice)
        out.loc = ItemGetter_loc(out)
        out.iloc = ItemGetter_iloc(out)
        return out

cumulative_dist_decoding_error(bst, *, tuningcurve, extern, decodefunc=decode1D, transfunc=None, n_bins=None)

Compute the cumulative distribution of decoding errors using a fixed tuning curve.

Parameters:

Name Type Description Default
bst BinnedSpikeTrainArray

Binned spike train array containing all epochs to decode.

 required
tuningcurve TuningCurve1D

Tuning curve to use for decoding.

 required
extern object

Query-able object of external correlates (e.g., position AnalogSignalArray).

 required
decodefunc callable

Decoding function to use (default is decode1D).

 decode1D
transfunc callable

Function to transform external variable (default is None).

 None
n_bins int

Number of decoding error bins (default is 200).

 None

Returns:

Name Type Description
cumhist Cumhist

Cumulative histogram of decoding errors.
bincenters ndarray

Bin centers for the cumulative histogram.
Source code in nelpy/decoding.py 
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
def cumulative_dist_decoding_error(
    bst, *, tuningcurve, extern, decodefunc=decode1D, transfunc=None, n_bins=None
):
    """
    Compute the cumulative distribution of decoding errors using a fixed tuning curve.

    Parameters
    ----------
    bst : BinnedSpikeTrainArray
        Binned spike train array containing all epochs to decode.
    tuningcurve : TuningCurve1D
        Tuning curve to use for decoding.
    extern : object
        Query-able object of external correlates (e.g., position AnalogSignalArray).
    decodefunc : callable, optional
        Decoding function to use (default is decode1D).
    transfunc : callable, optional
        Function to transform external variable (default is None).
    n_bins : int, optional
        Number of decoding error bins (default is 200).

    Returns
    -------
    cumhist : Cumhist
        Cumulative histogram of decoding errors.
    bincenters : np.ndarray
        Bin centers for the cumulative histogram.
    """

    def _trans_func(extern, at):
        """Default transform function to map extern into numerical bins"""

        _, ext = extern.asarray(at=at)

        return ext

    if transfunc is None:
        transfunc = _trans_func
    if n_bins is None:
        n_bins = 200

    # indices of training and validation epochs / events

    max_error = tuningcurve.bins[-1] - tuningcurve.bins[0]

    posterior, _, mode_pth, mean_pth = decodefunc(bst=bst, ratemap=tuningcurve)
    target = transfunc(extern, at=bst.bin_centers)
    hist, bins = np.histogram(
        np.abs(target - mean_pth), bins=n_bins, range=(0, max_error)
    )

    # build cumulative error distribution
    cumhist = np.cumsum(hist)
    cumhist = cumhist / cumhist[-1]
    bincenters = (bins + (bins[1] - bins[0]) / 2)[:-1]

    # modify to start at (0,0):
    cumhist = np.insert(cumhist, 0, 0)
    bincenters = np.insert(bincenters, 0, 0)

    # modify to end at (max_error,1):
    cumhist = np.append(cumhist, 1)
    bincenters = np.append(bincenters, max_error)

    cumhist = Cumhist(cumhist, bincenters)

    return cumhist, bincenters

cumulative_dist_decoding_error_using_xval(bst, extern, *, decodefunc=decode1D, k=5, transfunc=None, n_extern=100, extmin=0, extmax=100, sigma=3, n_bins=None, randomize=False)

Compute the cumulative distribution of decoding errors using k-fold cross-validation.

Parameters:

Name Type Description Default
bst BinnedSpikeTrainArray

Binned spike train array containing all epochs to decode.

 required
extern object

Query-able object of external correlates (e.g., position AnalogSignalArray).

 required
decodefunc callable

Decoding function to use (default is decode1D).

 decode1D
k int

Number of folds for cross-validation (default is 5).

 5
transfunc callable

Function to transform external variable (default is None).

 None
n_extern int

Number of external bins (default is 100).

 100
extmin float

Minimum value of external variable (default is 0).

 0
extmax float

Maximum value of external variable (default is 100).

 100
sigma float

Smoothing parameter for tuning curve (default is 3).

 3
n_bins int

Number of decoding error bins (default is 200).

 None
randomize bool

If True, randomize the order of epochs (default is False).

 False

Returns:

Name Type Description
cumhist Cumhist

Cumulative histogram of decoding errors.
bincenters ndarray

Bin centers for the cumulative histogram.
Source code in nelpy/decoding.py 
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
def cumulative_dist_decoding_error_using_xval(
    bst,
    extern,
    *,
    decodefunc=decode1D,
    k=5,
    transfunc=None,
    n_extern=100,
    extmin=0,
    extmax=100,
    sigma=3,
    n_bins=None,
    randomize=False,
):
    """
    Compute the cumulative distribution of decoding errors using k-fold cross-validation.

    Parameters
    ----------
    bst : BinnedSpikeTrainArray
        Binned spike train array containing all epochs to decode.
    extern : object
        Query-able object of external correlates (e.g., position AnalogSignalArray).
    decodefunc : callable, optional
        Decoding function to use (default is decode1D).
    k : int, optional
        Number of folds for cross-validation (default is 5).
    transfunc : callable, optional
        Function to transform external variable (default is None).
    n_extern : int, optional
        Number of external bins (default is 100).
    extmin : float, optional
        Minimum value of external variable (default is 0).
    extmax : float, optional
        Maximum value of external variable (default is 100).
    sigma : float, optional
        Smoothing parameter for tuning curve (default is 3).
    n_bins : int, optional
        Number of decoding error bins (default is 200).
    randomize : bool, optional
        If True, randomize the order of epochs (default is False).

    Returns
    -------
    cumhist : Cumhist
        Cumulative histogram of decoding errors.
    bincenters : np.ndarray
        Bin centers for the cumulative histogram.
    """

    def _trans_func(extern, at):
        """Default transform function to map extern into numerical bins"""

        _, ext = extern.asarray(at=at)

        return ext

    if transfunc is None:
        transfunc = _trans_func

    if n_bins is None:
        n_bins = 200

    max_error = extmax - extmin

    # indices of training and validation epochs / events

    hist = np.zeros(n_bins)
    for training, validation in k_fold_cross_validation(
        bst.n_epochs, k=k, randomize=randomize
    ):
        # estimate place fields using bst[training]
        tc = auxiliary.TuningCurve1D(
            bst=bst[training],
            extern=extern,
            n_extern=n_extern,
            extmin=extmin,
            extmax=extmax,
            sigma=sigma,
        )
        # decode position using bst[validation]
        posterior, _, mode_pth, mean_pth = decodefunc(bst[validation], tc)
        # calculate validation error (for current fold) by comapring
        # decoded pos v target pos
        target = transfunc(extern, at=bst[validation].bin_centers)

        histnew, bins = np.histogram(
            np.abs(target - mean_pth), bins=n_bins, range=(0, max_error)
        )
        hist = hist + histnew

    # build cumulative error distribution
    cumhist = np.cumsum(hist)
    cumhist = cumhist / cumhist[-1]
    bincenters = (bins + (bins[1] - bins[0]) / 2)[:-1]

    # modify to start at (0,0):
    cumhist = np.insert(cumhist, 0, 0)
    bincenters = np.insert(bincenters, 0, 0)

    # modify to end at (max_error,1):
    cumhist = np.append(cumhist, 1)
    bincenters = np.append(bincenters, max_error)

    cumhist = Cumhist(cumhist, bincenters)
    return cumhist, bincenters

decode1D(bst, ratemap, xmin=0, xmax=100, w=1, nospk_prior=None, _skip_empty_bins=True)

Decode binned spike trains using a 1D ratemap (Bayesian decoding).

Parameters:

Name Type Description Default
bst BinnedSpikeTrainArray

Binned spike train array to decode.

 required
ratemap array_like or TuningCurve1D

Firing rate map with shape (n_units, n_ext), where n_ext is the number of external correlates (e.g., position bins). The rate map is in spks/second.

 required
xmin float

Minimum value of external variable (default is 0).

 0
xmax float

Maximum value of external variable (default is 100).

 100
w int

Window size for decoding (default is 1).

 1
nospk_prior array_like or float

Prior distribution over external correlates with shape (n_ext,). Used if no spikes are observed in a decoding window. If scalar, a uniform prior is assumed. Default is np.nan.

 None
_skip_empty_bins bool

If True, skip bins with no spikes. If False, fill with prior.

 True

Returns:

Name Type Description
posterior ndarray

Posterior distribution with shape (n_ext, n_posterior_bins).
cum_posterior_lengths ndarray

Cumulative posterior lengths for each epoch.
mode_pth ndarray

Most likely position at each time bin.
mean_pth ndarray

Expected position at each time bin.

Examples:

>>> posterior, cum_posterior_lengths, mode_pth, mean_pth = decode1D(bst, ratemap)
Source code in nelpy/decoding.py 
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
def decode1D(
    bst, ratemap, xmin=0, xmax=100, w=1, nospk_prior=None, _skip_empty_bins=True
):
    """
    Decode binned spike trains using a 1D ratemap (Bayesian decoding).

    Parameters
    ----------
    bst : BinnedSpikeTrainArray
        Binned spike train array to decode.
    ratemap : array_like or TuningCurve1D
        Firing rate map with shape (n_units, n_ext), where n_ext is the number of external correlates (e.g., position bins). The rate map is in spks/second.
    xmin : float, optional
        Minimum value of external variable (default is 0).
    xmax : float, optional
        Maximum value of external variable (default is 100).
    w : int, optional
        Window size for decoding (default is 1).
    nospk_prior : array_like or float, optional
        Prior distribution over external correlates with shape (n_ext,). Used if no spikes are observed in a decoding window. If scalar, a uniform prior is assumed. Default is np.nan.
    _skip_empty_bins : bool, optional
        If True, skip bins with no spikes. If False, fill with prior.

    Returns
    -------
    posterior : np.ndarray
        Posterior distribution with shape (n_ext, n_posterior_bins).
    cum_posterior_lengths : np.ndarray
        Cumulative posterior lengths for each epoch.
    mode_pth : np.ndarray
        Most likely position at each time bin.
    mean_pth : np.ndarray
        Expected position at each time bin.

    Examples
    --------
    >>> posterior, cum_posterior_lengths, mode_pth, mean_pth = decode1D(bst, ratemap)
    """

    if w is None:
        w = 1
    assert float(w).is_integer(), "w must be a positive integer!"
    assert w > 0, "w must be a positive integer!"

    n_units, t_bins = bst.data.shape
    _, n_xbins = ratemap.shape

    # if we pass a TuningCurve1D object, extract the ratemap and re-order
    # units if necessary
    if isinstance(ratemap, auxiliary.TuningCurve1D) | isinstance(
        ratemap, auxiliary._tuningcurve.TuningCurve1D
    ):
        # xmin = ratemap.bins[0]
        xmax = ratemap.bins[-1]
        bin_centers = ratemap.bin_centers
        # re-order units if necessary
        ratemap = ratemap.reorder_units_by_ids(bst.unit_ids)
        ratemap = ratemap.ratemap
    else:
        # xmin = 0
        xmax = n_xbins
        bin_centers = np.arange(n_xbins)

    if nospk_prior is None:
        nospk_prior = np.full(n_xbins, np.nan)
    elif isinstance(nospk_prior, numbers.Number):
        nospk_prior = np.full(n_xbins, 1.0)

    assert nospk_prior.shape[0] == n_xbins, "prior must have length {}".format(n_xbins)
    assert nospk_prior.size == n_xbins, (
        "prior must be a 1D array with length {}".format(n_xbins)
    )

    lfx = np.log(ratemap)

    eterm = -ratemap.sum(axis=0) * bst.ds * w

    # if we decode using multiple bins at a time (w>1) then we have to decode each epoch separately:

    # first, we determine the number of bins we will decode. This requires us to scan over the epochs
    n_bins = 0
    cumlengths = np.cumsum(bst.lengths)
    posterior_lengths = np.zeros(bst.n_epochs, dtype=int)
    prev_idx = 0
    for ii, to_idx in enumerate(cumlengths):
        datalen = to_idx - prev_idx
        prev_idx = to_idx
        posterior_lengths[ii] = np.max((1, datalen - w + 1))

    n_bins = posterior_lengths.sum()
    posterior = np.zeros((n_xbins, n_bins))

    # next, we decode each epoch separately, one bin at a time
    cum_posterior_lengths = np.insert(np.cumsum(posterior_lengths), 0, 0)
    prev_idx = 0
    for ii, to_idx in enumerate(cumlengths):
        data = bst.data[:, prev_idx:to_idx]
        prev_idx = to_idx
        datacum = np.cumsum(
            data, axis=1
        )  # ii'th data segment, with column of zeros prepended
        datacum = np.hstack((np.zeros((n_units, 1)), datacum))
        re = w  # right edge ptr
        # TODO: check if datalen < w and act appropriately
        if posterior_lengths[ii] > 1:  # more than one full window fits into data length
            for tt in range(posterior_lengths[ii]):
                obs = datacum[:, re] - datacum[:, re - w]  # spikes in window of size w
                re += 1
                post_idx = cum_posterior_lengths[ii] + tt
                if obs.sum() == 0 and _skip_empty_bins:
                    # no spikes to decode in window!
                    posterior[:, post_idx] = nospk_prior
                else:
                    posterior[:, post_idx] = (
                        np.tile(np.array(obs, ndmin=2).T, n_xbins) * lfx
                    ).sum(axis=0) + eterm
        else:  # only one window can fit in, and perhaps only partially. We just take all the data we can get,
            # and ignore the scaling problem where the window size is now possibly less than bst.ds*w
            post_idx = cum_posterior_lengths[ii]
            obs = datacum[:, -1]  # spikes in window of size at most w
            if obs.sum() == 0 and _skip_empty_bins:
                # no spikes to decode in window!
                posterior[:, post_idx] = nospk_prior
            else:
                posterior[:, post_idx] = (
                    np.tile(np.array(obs, ndmin=2).T, n_xbins) * lfx
                ).sum(axis=0) + eterm

    # normalize posterior:
    posterior = np.exp(posterior - logsumexp(posterior, axis=0))

    # TODO: what was my rationale behid the following? Why not use bin centers?
    # _, bins = np.histogram([], bins=n_xbins, range=(xmin,xmax))
    # xbins = (bins + xmax/n_xbins)[:-1]

    mode_pth = np.argmax(posterior, axis=0) * xmax / n_xbins
    mode_pth = np.where(np.isnan(posterior.sum(axis=0)), np.nan, mode_pth)
    mean_pth = (bin_centers * posterior.T).sum(axis=1)
    return posterior, cum_posterior_lengths, mode_pth, mean_pth

decode2D(bst, ratemap, xmin=0, xmax=100, ymin=0, ymax=100, w=1, nospk_prior=None, _skip_empty_bins=True)

Decode binned spike trains using a 2D ratemap (Bayesian decoding).

Parameters:

Name Type Description Default
bst BinnedSpikeTrainArray

Binned spike train array to decode.

 required
ratemap array_like or TuningCurve2D

Firing rate map with shape (n_units, ext_nx, ext_ny), where ext_nx and ext_ny are the number of external correlates (e.g., position bins). The rate map is in spks/second.

 required
xmin float

Minimum x value of external variable (default is 0).

 0
xmax float

Maximum x value of external variable (default is 100).

 100
ymin float

Minimum y value of external variable (default is 0).

 0
ymax float

Maximum y value of external variable (default is 100).

 100
w int

Window size for decoding (default is 1).

 1
nospk_prior array_like or float

Prior distribution over external correlates with shape (ext_nx, ext_ny). Used if no spikes are observed in a decoding window. If scalar, a uniform prior is assumed. Default is np.nan.

 None
_skip_empty_bins bool

If True, skip bins with no spikes. If False, fill with prior.

 True

Returns:

Name Type Description
posterior ndarray

Posterior distribution with shape (ext_nx, ext_ny, n_posterior_bins).
cum_posterior_lengths ndarray

Cumulative posterior lengths for each epoch.
mode_pth ndarray

Most likely (x, y) position at each time bin.
mean_pth ndarray

Expected (x, y) position at each time bin.

Examples:

>>> posterior, cum_posterior_lengths, mode_pth, mean_pth = decode2D(bst, ratemap)
Source code in nelpy/decoding.py 
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
def decode2D(
    bst,
    ratemap,
    xmin=0,
    xmax=100,
    ymin=0,
    ymax=100,
    w=1,
    nospk_prior=None,
    _skip_empty_bins=True,
):
    """
    Decode binned spike trains using a 2D ratemap (Bayesian decoding).

    Parameters
    ----------
    bst : BinnedSpikeTrainArray
        Binned spike train array to decode.
    ratemap : array_like or TuningCurve2D
        Firing rate map with shape (n_units, ext_nx, ext_ny), where ext_nx and ext_ny are the number of external correlates (e.g., position bins). The rate map is in spks/second.
    xmin : float, optional
        Minimum x value of external variable (default is 0).
    xmax : float, optional
        Maximum x value of external variable (default is 100).
    ymin : float, optional
        Minimum y value of external variable (default is 0).
    ymax : float, optional
        Maximum y value of external variable (default is 100).
    w : int, optional
        Window size for decoding (default is 1).
    nospk_prior : array_like or float, optional
        Prior distribution over external correlates with shape (ext_nx, ext_ny). Used if no spikes are observed in a decoding window. If scalar, a uniform prior is assumed. Default is np.nan.
    _skip_empty_bins : bool, optional
        If True, skip bins with no spikes. If False, fill with prior.

    Returns
    -------
    posterior : np.ndarray
        Posterior distribution with shape (ext_nx, ext_ny, n_posterior_bins).
    cum_posterior_lengths : np.ndarray
        Cumulative posterior lengths for each epoch.
    mode_pth : np.ndarray
        Most likely (x, y) position at each time bin.
    mean_pth : np.ndarray
        Expected (x, y) position at each time bin.

    Examples
    --------
    >>> posterior, cum_posterior_lengths, mode_pth, mean_pth = decode2D(bst, ratemap)
    """

    def tile_obs(obs, nx, ny):
        n_units = len(obs)
        out = np.zeros((n_units, nx, ny))
        for unit in range(n_units):
            out[unit, :, :] = obs[unit]
        return out

    if w is None:
        w = 1
    assert float(w).is_integer(), "w must be a positive integer!"
    assert w > 0, "w must be a positive integer!"

    n_units, t_bins = bst.data.shape

    xbins = None
    ybins = None

    # if we pass a TuningCurve2D object, extract the ratemap and re-order
    # units if necessary
    if isinstance(ratemap, auxiliary.TuningCurve2D):
        xbins = ratemap.xbins
        ybins = ratemap.ybins
        xbin_centers = ratemap.xbin_centers
        ybin_centers = ratemap.ybin_centers
        # re-order units if necessary
        ratemap = ratemap.reorder_units_by_ids(bst.unit_ids)
        ratemap = ratemap.ratemap

    _, n_xbins, n_ybins = ratemap.shape

    if nospk_prior is None:
        nospk_prior = np.full((n_xbins, n_ybins), np.nan)
    elif isinstance(nospk_prior, numbers.Number):
        nospk_prior = np.full((n_xbins, n_ybins), 1.0)

    assert nospk_prior.shape == (
        n_xbins,
        n_ybins,
    ), "prior must have shape ({}, {})".format(n_xbins, n_ybins)

    lfx = np.log(ratemap)

    eterm = -ratemap.sum(axis=0) * bst.ds * w

    # if we decode using multiple bins at a time (w>1) then we have to decode each epoch separately:

    # first, we determine the number of bins we will decode. This requires us to scan over the epochs
    n_tbins = 0
    cumlengths = np.cumsum(bst.lengths)
    posterior_lengths = np.zeros(bst.n_epochs, dtype=int)
    prev_idx = 0
    for ii, to_idx in enumerate(cumlengths):
        datalen = to_idx - prev_idx
        prev_idx = to_idx
        posterior_lengths[ii] = np.max((1, datalen - w + 1))

    n_tbins = posterior_lengths.sum()

    ########################################################################
    posterior = np.zeros((n_xbins, n_ybins, n_tbins))

    # next, we decode each epoch separately, one bin at a time
    cum_posterior_lengths = np.insert(np.cumsum(posterior_lengths), 0, 0)
    prev_idx = 0
    for ii, to_idx in enumerate(cumlengths):
        data = bst.data[:, prev_idx:to_idx]
        prev_idx = to_idx
        datacum = np.cumsum(
            data, axis=1
        )  # ii'th data segment, with column of zeros prepended
        datacum = np.hstack((np.zeros((n_units, 1)), datacum))
        re = w  # right edge ptr
        # TODO: check if datalen < w and act appropriately
        if posterior_lengths[ii] > 1:  # more than one full window fits into data length
            for tt in range(posterior_lengths[ii]):
                obs = datacum[:, re] - datacum[:, re - w]  # spikes in window of size w
                re += 1
                post_idx = cum_posterior_lengths[ii] + tt
                if obs.sum() == 0 and not _skip_empty_bins:
                    # no spikes to decode in window!
                    posterior[:, :, post_idx] = nospk_prior
                else:
                    posterior[:, :, post_idx] = (
                        tile_obs(obs, n_xbins, n_ybins) * lfx
                    ).sum(axis=0) + eterm
        else:  # only one window can fit in, and perhaps only partially. We just take all the data we can get,
            # and ignore the scaling problem where the window size is now possibly less than bst.ds*w
            post_idx = cum_posterior_lengths[ii]
            obs = datacum[:, -1]  # spikes in window of size at most w
            if obs.sum() == 0 and not _skip_empty_bins:
                # no spikes to decode in window!
                posterior[:, :, post_idx] = nospk_prior
            else:
                posterior[:, :, post_idx] = (tile_obs(obs, n_xbins, n_ybins) * lfx).sum(
                    axis=0
                ) + eterm

    # normalize posterior:
    # see http://timvieira.github.io/blog/post/2014/02/11/exp-normalize-trick/

    for tt in range(n_tbins):
        posterior[:, :, tt] = posterior[:, :, tt] - posterior[:, :, tt].max()
        posterior[:, :, tt] = np.exp(posterior[:, :, tt])
        posterior[:, :, tt] = posterior[:, :, tt] / posterior[:, :, tt].sum()

    # if xbins is None:
    #     _, bins = np.histogram([], bins=n_xbins, range=(xmin,xmax))
    #     xbins = (bins + xmax/n_xbins)[:-1]
    # if ybins is None:
    #     _, bins = np.histogram([], bins=n_ybins, range=(ymin,ymax))
    #     ybins = (bins + ymax/n_ybins)[:-1]

    mode_pth = np.zeros((2, n_tbins))
    for tt in range(n_tbins):
        if np.any(np.isnan(posterior[:, :, tt])):
            mode_pth[0, tt] = np.nan
            mode_pth[0, tt] = np.nan
        else:
            x_, y_ = np.unravel_index(
                np.argmax(posterior[:, :, tt]), (n_xbins, n_ybins)
            )
            mode_pth[0, tt] = xbins[x_]
            mode_pth[1, tt] = ybins[y_]

    expected_x = (xbin_centers * posterior.sum(axis=1).T).sum(axis=1)
    expected_y = (ybin_centers * posterior.sum(axis=0).T).sum(axis=1)
    mean_pth = np.vstack((expected_x, expected_y))

    posterior = np.transpose(posterior, axes=[1, 0, 2])

    return posterior, cum_posterior_lengths, mode_pth, mean_pth

get_mean_pth_from_array(posterior, tuningcurve=None)

Compute the mean path (expected position) from a posterior probability matrix.

If a tuning curve is provided, the mean is mapped back to external coordinates/units. Otherwise, the mean is in bin space.

Parameters:

Name Type Description Default
posterior ndarray

Posterior probability matrix (position x time).

 required
tuningcurve TuningCurve1D

Tuning curve for mapping bins to external coordinates.

 None

Returns:

Name Type Description
mean_pth ndarray

Expected position at each time bin.
Source code in nelpy/decoding.py 
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
def get_mean_pth_from_array(posterior, tuningcurve=None):
    """
    Compute the mean path (expected position) from a posterior probability matrix.

    If a tuning curve is provided, the mean is mapped back to external coordinates/units.
    Otherwise, the mean is in bin space.

    Parameters
    ----------
    posterior : np.ndarray
        Posterior probability matrix (position x time).
    tuningcurve : TuningCurve1D, optional
        Tuning curve for mapping bins to external coordinates.

    Returns
    -------
    mean_pth : np.ndarray
        Expected position at each time bin.
    """
    n_xbins = posterior.shape[0]

    if tuningcurve is None:
        xmin = 0
        xmax = 1
    else:
        # TODO: this only works for TuningCurve1D currently
        if isinstance(tuningcurve, auxiliary.TuningCurve1D):
            xmin = tuningcurve.bins[0]
            xmax = tuningcurve.bins[-1]
        else:
            raise TypeError("tuningcurve type not yet supported!")

    _, bins = np.histogram([], bins=n_xbins, range=(xmin, xmax))
    xbins = (bins + xmax / n_xbins)[:-1]

    mean_pth = (xbins * posterior.T).sum(axis=1)

    return mean_pth

get_mode_pth_from_array(posterior, tuningcurve=None)

Compute the mode path (most likely position) from a posterior probability matrix.

If a tuning curve is provided, the mode is mapped back to external coordinates/units. Otherwise, the mode is in bin space.

Parameters:

Name Type Description Default
posterior ndarray

Posterior probability matrix (position x time).

 required
tuningcurve TuningCurve1D

Tuning curve for mapping bins to external coordinates.

 None

Returns:

Name Type Description
mode_pth ndarray

Most likely position at each time bin.
Source code in nelpy/decoding.py 
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
def get_mode_pth_from_array(posterior, tuningcurve=None):
    """
    Compute the mode path (most likely position) from a posterior probability matrix.

    If a tuning curve is provided, the mode is mapped back to external coordinates/units.
    Otherwise, the mode is in bin space.

    Parameters
    ----------
    posterior : np.ndarray
        Posterior probability matrix (position x time).
    tuningcurve : TuningCurve1D, optional
        Tuning curve for mapping bins to external coordinates.

    Returns
    -------
    mode_pth : np.ndarray
        Most likely position at each time bin.
    """
    n_xbins = posterior.shape[0]

    if tuningcurve is None:
        xmin = 0
        xmax = n_xbins
    else:
        # TODO: this only works for TuningCurve1D currently
        if isinstance(tuningcurve, auxiliary.TuningCurve1D):
            xmin = tuningcurve.bins[0]
            xmax = tuningcurve.bins[-1]
        else:
            raise TypeError("tuningcurve type not yet supported!")

    _, bins = np.histogram([], bins=n_xbins, range=(xmin, xmax))
    # xbins = (bins + xmax / n_xbins)[:-1]

    mode_pth = np.argmax(posterior, axis=0) * xmax / n_xbins
    mode_pth = np.where(np.isnan(posterior.sum(axis=0)), np.nan, mode_pth)

    return mode_pth

k_fold_cross_validation(X, k=None, randomize=False)

Generate K (training, validation) pairs from the items in X for cross-validation.

Parameters:

Name Type Description Default
X list or int

List of items, list of indices, or integer number of indices.

 required
k int or str

Number of folds for k-fold cross-validation. 'loo' or 'LOO' for leave-one-out. Default is 5.

 None
randomize bool

If True, shuffle X before partitioning. Default is False.

 False

Yields:

Name Type Description
training list

Training set indices.
validation list

Validation set indices.

Examples:

>>> X = [i for i in range(97)]
>>> for training, validation in k_fold_cross_validation(X, k=5):
...     print(training, validation)
Source code in nelpy/decoding.py 
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
def k_fold_cross_validation(X, k=None, randomize=False):
    """
    Generate K (training, validation) pairs from the items in X for cross-validation.

    Parameters
    ----------
    X : list or int
        List of items, list of indices, or integer number of indices.
    k : int or str, optional
        Number of folds for k-fold cross-validation. 'loo' or 'LOO' for leave-one-out. Default is 5.
    randomize : bool, optional
        If True, shuffle X before partitioning. Default is False.

    Yields
    ------
    training : list
        Training set indices.
    validation : list
        Validation set indices.

    Examples
    --------
    >>> X = [i for i in range(97)]
    >>> for training, validation in k_fold_cross_validation(X, k=5):
    ...     print(training, validation)
    """
    # deal with default values:
    if isinstance(X, int):
        X = range(X)
    n_samples = len(X)
    if k is None:
        k = 5
    elif k == "loo" or k == "LOO":
        k = n_samples

    if randomize:
        from random import shuffle

        X = list(X)
        shuffle(X)
    for _k_ in range(k):
        training = [x for i, x in enumerate(X) if i % k != _k_]
        validation = [x for i, x in enumerate(X) if i % k == _k_]
        try:
            yield training, validation
        except StopIteration:
            return

rmse(predictions, targets)

Calculate the root mean squared error (RMSE) between predictions and targets.

Parameters:

Name Type Description Default
predictions array_like

Array of predicted values.

 required
targets array_like

Array of target values.

 required

Returns:

Name Type Description
rmse float

Root mean squared error of the predictions with respect to the targets.
Source code in nelpy/decoding.py 
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
def rmse(predictions, targets):
    """
    Calculate the root mean squared error (RMSE) between predictions and targets.

    Parameters
    ----------
    predictions : array_like
        Array of predicted values.
    targets : array_like
        Array of target values.

    Returns
    -------
    rmse : float
        Root mean squared error of the predictions with respect to the targets.
    """
    predictions = np.asanyarray(predictions)
    targets = np.asanyarray(targets)
    rmse = np.sqrt(np.nanmean((predictions - targets) ** 2))
    return rmse