maybe_unsqueeze[source]

maybe_unsqueeze(x)

Add empty dimension if it is a rank 1 tensor/array

a = np.random.random(10)
test_eq((1,10), maybe_unsqueeze(a).shape)
test_eq((1,10), maybe_unsqueeze(maybe_unsqueeze(a)).shape) #do nothing

t = torch.rand(10)
test_eq((1,10), maybe_unsqueeze(t).shape)
test_eq((1,10), maybe_unsqueeze(maybe_unsqueeze(t)).shape) #do nothing

A time series is just an array of 1 dimesion.

show_array[source]

show_array(array, ax=None, figsize=None, title=None, ctx=None, tx=None, **kwargs)

Show an array on ax.

A simple array of 1 channel is np.arange(10).

show_array(np.arange(10));

class TSeries[source]

TSeries(x, **kwargs) :: TensorBase

Basic Timeseries wrapper

We can add some noise for the fun

a = np.arange(10)+np.random.randn(10)

a.shape

(10,)

As we want to make explicit that is a one channel timeseries, we will unsqueeze the first dimension.

ts = TSeries.create(a)

ts.data

tensor([[-1.0373,  2.2527,  2.2667,  4.0746,  4.2688,  4.9019,  3.6700,  7.7529,
          6.5165,  8.9746]])

ts.shape

torch.Size([1, 10])

ts.show();

ts.data.mean()

tensor(4.3641)

@patch
def encodes(self:Normalize, x:TSeries): return (x - self.mean) / self.std

@patch
def encodes(self: Normalize, x:TSeries): 
    f = to_cpu if x.device.type=='cpu' else noop
    return (x*f(self.std) + f(self.mean))

nrm = Normalize(ts.data.mean(), ts.data.std())

((ts-nrm.mean)/ts.data.std()).show()

<AxesSubplot:>

nrm.encodes(ts).show();

UCR

The 2018 UCR 128-timeseries dataset

This is the dataset used to benchmark algorithms for Timeseries classification.

URLs.UCR

'http://www.timeseriesclassification.com/Downloads/Archives/Univariate2018_arff.zip'

ucr_path = untar_data(URLs.UCR)

Each sub task is on a subfolde with task name. For instance we will select 'Adiac' .

ucr_path.ls()

(#135) [Path('/home/tcapelle/.fastai/data/Univariate2018_arff/GestureMidAirD2'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/Adiac'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/CinCECGTorso'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/LargeKitchenAppliances'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/MixedShapesRegularTrain'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/ShakeGestureWiimoteZ'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/SmallKitchenAppliances'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/CricketY'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/ToeSegmentation2'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/EOGVerticalSignal')...]

adiac_path = ucr_path/'Adiac'

adiac_path.ls()

(#5) [Path('/home/tcapelle/.fastai/data/Univariate2018_arff/Adiac/Adiac_TEST.txt'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/Adiac/Adiac_TEST.arff'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/Adiac/Adiac_TRAIN.arff'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/Adiac/Adiac_TRAIN.txt'),Path('/home/tcapelle/.fastai/data/Univariate2018_arff/Adiac/Adiac.txt')]

We can find .csv and .arff files on this copy of the dataset. We will read the .arff files using a helper function. We will store them on pandas DataFrame's

load_df_ucr[source]

load_df_ucr(path, task)

Loads arff files from UCR pandas dfs

df_train, df_test = load_df_ucr(ucr_path, 'StarLightCurves')

Loading files from: /home/tcapelle/.fastai/data/Univariate2018_arff/StarLightCurves

df_train.head()

	att1	att2	att3	att4	att5	att6	att7	att8	att9	att10	...	att1016	att1017	att1018	att1019	att1020	att1021	att1022	att1023	att1024	target
0	0.537303	0.531103	0.528503	0.529403	0.533603	0.540903	0.551103	0.564003	0.579603	0.597603	...	0.546903	0.545903	0.543903	0.541003	0.537203	0.532303	0.526403	0.519503	0.511403	b'3'
1	0.588398	0.593898	0.599098	0.604098	0.608798	0.613397	0.617797	0.622097	0.626097	0.630097	...	0.237399	0.246499	0.256199	0.266499	0.277399	0.288799	0.300899	0.313599	0.326899	b'3'
2	-0.049900	-0.041500	-0.033400	-0.025600	-0.018100	-0.010800	-0.003800	0.003000	0.009600	0.015900	...	-0.173801	-0.161601	-0.149201	-0.136401	-0.123201	-0.109701	-0.095901	-0.081701	-0.067100	b'1'
3	1.337005	1.319805	1.302905	1.286305	1.270005	1.254005	1.238304	1.223005	1.208104	1.193504	...	1.288905	1.298505	1.307705	1.316505	1.324905	1.332805	1.340205	1.347005	1.353205	b'3'
4	0.769801	0.775301	0.780401	0.785101	0.789401	0.793301	0.796801	0.799901	0.802601	0.805101	...	0.742401	0.744501	0.747301	0.750701	0.754801	0.759501	0.765001	0.771301	0.778401	b'3'

5 rows × 1025 columns

load_np_ucr[source]

load_np_ucr(path, task)

Loads arff files from UCR into np arrays

x_train, y_train, x_test, y_test = load_np_ucr(ucr_path, 'StarLightCurves')

Loading files from: /home/tcapelle/.fastai/data/Univariate2018_arff/StarLightCurves

test_eq(len(x_train), len(y_train))
test_eq(len(x_test), len(y_test))

Datasets

Loading from DataFrames

def get_x(row):
    return row.values[:-1].astype('float32')
def get_y(row):
    return int(row.values[-1])

x_cols = df_train.columns[slice(0,-1)].to_list()
x_cols[0:5]

['att1', 'att2', 'att3', 'att4', 'att5']

TSeries(get_x(df_train.iloc[0])).show();

splits = [list(range(len(df_train))), list(range(len(df_train), len(df_train)+len(df_test)))]

ds = Datasets(pd.concat([df_train, df_test]).reset_index(drop=True), 
                    tfms=[[get_x, TSeries.create], [get_y, Categorize()]],
                    splits=splits, 
                   )

ds.valid

(#8236) [(TSeries(ch=1, len=1024), TensorCategory(1)),(TSeries(ch=1, len=1024), TensorCategory(2)),(TSeries(ch=1, len=1024), TensorCategory(2)),(TSeries(ch=1, len=1024), TensorCategory(0)),(TSeries(ch=1, len=1024), TensorCategory(1)),(TSeries(ch=1, len=1024), TensorCategory(2)),(TSeries(ch=1, len=1024), TensorCategory(2)),(TSeries(ch=1, len=1024), TensorCategory(2)),(TSeries(ch=1, len=1024), TensorCategory(1)),(TSeries(ch=1, len=1024), TensorCategory(2))...]

dls = ds.dataloaders(bs=2)

dls.show_batch()