# Run this cell to install the latest version of fastcore shared on github
!pip install git+https://github.com/fastai/fastai2.git

# Run this cell to install the latest version of fastcore shared on github
!pip install git+https://github.com/fastai/fastcore.git

# Run this cell to install the latest version of timeseries shared on github
!pip install git+https://github.com/ai-fast-track/timeseries.git

%reload_ext autoreload
%autoreload 2
%matplotlib inline

from fastai2.basics import *

from timeseries.all import *

Yoga Dataset

The dataset was obtained by capturing two actors transiting between yoga poses in front of a green screen. The problem is to discriminate between one actor (male) and another (female). Each image was converted to a one dimensional series by finding the outline and measuring the distance of the outline to the centre.

dsname =  'Yoga' # 'ChlorineConcentration', 'Yoga'

url = 'http://www.timeseriesclassification.com/Downloads/Yoga.zip'
path = unzip_data(URLs_TS.UNI_YOGA)
path

Path('/home/farid/.fastai/data/Yoga')

fname_train = f'{dsname}_TRAIN.arff'
fname_test = f'{dsname}_TEST.arff'
fnames = [path/fname_train, path/fname_test]
fnames

[Path('/home/farid/.fastai/data/Yoga/Yoga_TRAIN.arff'),
 Path('/home/farid/.fastai/data/Yoga/Yoga_TEST.arff')]

data = TSData.from_arff(fnames)
print(data)

TSData:
 Datasets names (concatenated): ['Yoga_TRAIN', 'Yoga_TEST']
 Filenames:                     [Path('/home/farid/.fastai/data/Yoga/Yoga_TRAIN.arff'), Path('/home/farid/.fastai/data/Yoga/Yoga_TEST.arff')]
 Data shape: (3300, 1, 426)
 Targets shape: (3300,)
 Nb Samples: 3300
 Nb Channels:           1
 Sequence Length: 426

items = data.get_items()

seed = 42
splits = RandomSplitter(seed=seed)(range_of(items)) #by default 80% for train split and 20% for valid split are chosen 
splits

((#2640) [1683,2715,2242,824,768,1022,2227,1359,59,1381...],
 (#660) [3042,387,98,1891,430,2595,1282,2909,1774,1682...])

bs = 128                            
# Normalize at batch time
tfm_norm = Normalize(scale_subtype = 'per_sample_per_channel', scale_range=(0, 1)) # per_sample , per_sample_per_channel
# tfm_norm = Standardize(scale_subtype = 'per_sample')
batch_tfms = [tfm_norm]

default_device()

device(type='cpu')

# dls = TSDataLoaders.from_files(fnames=fnames, batch_tfms=batch_tfms, num_workers=0, device=default_device())
dls = TSDataLoaders.from_files(fnames=fnames, num_workers=0, device=default_device())

dls.show_batch(max_n=9)

Training Model

# Number of channels (i.e. dimensions in ARFF and TS files jargon)
c_in = get_n_channels(dls.train) # data.n_channels
# Number of classes
c_out= dls.c 
c_in,c_out

(1, 2)

model = inception_time(c_in, c_out).to(device=default_device())
model

Sequential(
  (0): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (convs): ModuleList(
          (0): Conv1d(1, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(1, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(1, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (maxpool_bottleneck): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(1, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (1): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,))
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (maxpool_bottleneck): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (2): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,))
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (maxpool_bottleneck): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
      (1): Shortcut(
        (act_fn): ReLU(inplace=True)
        (conv): Conv1d(128, 128, kernel_size=(1,), stride=(1,), bias=False)
        (bn): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
  )
  (3): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,))
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (maxpool_bottleneck): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (4): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,))
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (maxpool_bottleneck): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
    )
  )
  (5): SequentialEx(
    (layers): ModuleList(
      (0): InceptionModule(
        (bottleneck): Conv1d(128, 32, kernel_size=(1,), stride=(1,))
        (convs): ModuleList(
          (0): Conv1d(32, 32, kernel_size=(39,), stride=(1,), padding=(19,), bias=False)
          (1): Conv1d(32, 32, kernel_size=(19,), stride=(1,), padding=(9,), bias=False)
          (2): Conv1d(32, 32, kernel_size=(9,), stride=(1,), padding=(4,), bias=False)
        )
        (maxpool_bottleneck): Sequential(
          (0): MaxPool1d(kernel_size=3, stride=1, padding=1, dilation=1, ceil_mode=False)
          (1): Conv1d(128, 32, kernel_size=(1,), stride=(1,), bias=False)
        )
        (bn_relu): Sequential(
          (0): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (1): ReLU()
        )
      )
      (1): Shortcut(
        (act_fn): ReLU(inplace=True)
        (conv): Conv1d(128, 128, kernel_size=(1,), stride=(1,), bias=False)
        (bn): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
  )
  (6): AdaptiveConcatPool1d(
    (ap): AdaptiveAvgPool1d(output_size=1)
    (mp): AdaptiveMaxPool1d(output_size=1)
  )
  (7): Flatten(full=False)
  (8): Linear(in_features=256, out_features=2, bias=True)
)

# opt_func = partial(Adam, lr=3e-3, wd=0.01)
#Or use Ranger
def opt_func(p, lr=slice(3e-3)): return Lookahead(RAdam(p, lr=lr, mom=0.95, wd=0.01))

#Learner
loss_func = LabelSmoothingCrossEntropy() 
learn = Learner(dls, model, opt_func=opt_func, loss_func=loss_func, metrics=accuracy)

print(learn.summary())

Sequential (Input shape: ['64 x 1 x 426'])
================================================================
Layer (type)         Output Shape         Param #    Trainable 
================================================================
Conv1d               64 x 32 x 426        1,248      True      
________________________________________________________________
Conv1d               64 x 32 x 426        608        True      
________________________________________________________________
Conv1d               64 x 32 x 426        288        True      
________________________________________________________________
MaxPool1d            64 x 1 x 426         0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        32         True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,128      True      
________________________________________________________________
Conv1d               64 x 32 x 426        39,936     True      
________________________________________________________________
Conv1d               64 x 32 x 426        19,456     True      
________________________________________________________________
Conv1d               64 x 32 x 426        9,216      True      
________________________________________________________________
MaxPool1d            64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,096      True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,128      True      
________________________________________________________________
Conv1d               64 x 32 x 426        39,936     True      
________________________________________________________________
Conv1d               64 x 32 x 426        19,456     True      
________________________________________________________________
Conv1d               64 x 32 x 426        9,216      True      
________________________________________________________________
MaxPool1d            64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,096      True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 128 x 426       16,384     True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
Conv1d               64 x 32 x 426        4,128      True      
________________________________________________________________
Conv1d               64 x 32 x 426        39,936     True      
________________________________________________________________
Conv1d               64 x 32 x 426        19,456     True      
________________________________________________________________
Conv1d               64 x 32 x 426        9,216      True      
________________________________________________________________
MaxPool1d            64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,096      True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,128      True      
________________________________________________________________
Conv1d               64 x 32 x 426        39,936     True      
________________________________________________________________
Conv1d               64 x 32 x 426        19,456     True      
________________________________________________________________
Conv1d               64 x 32 x 426        9,216      True      
________________________________________________________________
MaxPool1d            64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,096      True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,128      True      
________________________________________________________________
Conv1d               64 x 32 x 426        39,936     True      
________________________________________________________________
Conv1d               64 x 32 x 426        19,456     True      
________________________________________________________________
Conv1d               64 x 32 x 426        9,216      True      
________________________________________________________________
MaxPool1d            64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 32 x 426        4,096      True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
ReLU                 64 x 128 x 426       0          False     
________________________________________________________________
Conv1d               64 x 128 x 426       16,384     True      
________________________________________________________________
BatchNorm1d          64 x 128 x 426       256        True      
________________________________________________________________
AdaptiveAvgPool1d    64 x 128 x 1         0          False     
________________________________________________________________
AdaptiveMaxPool1d    64 x 128 x 1         0          False     
________________________________________________________________
Flatten              64 x 256             0          False     
________________________________________________________________
Linear               64 x 2               514        True      
________________________________________________________________

Total params: 421,666
Total trainable params: 421,666
Total non-trainable params: 0

Optimizer used: <function opt_func at 0x7fdec02bb6a8>
Loss function: LabelSmoothingCrossEntropy()

Callbacks:
  - TrainEvalCallback
  - Recorder
  - ProgressCallback

lr_min, lr_steep = learn.lr_find()
lr_min, lr_steep

(0.014454397559165954, 4.3651581904669e-07)

lr_max=1e-3 # lr_max=lr_steep
epochs=10; lr_max=lr_max;  pct_start=.7; moms=(0.95,0.85,0.95); wd=1e-2
learn.fit_one_cycle(epochs, lr_max=lr_max, pct_start=pct_start,  moms=moms, wd=wd)
# learn.fit_one_cycle(epochs, lr_max=lr_steep)

Graphs

learn.recorder.plot_loss()

learn.show_results(max_n=9)

interp = ClassificationInterpretation.from_learner(learn)
interp.plot_confusion_matrix()

interp.most_confused()

[('1', '2', 19), ('2', '1', 4)]

Photo Credit: Farsai Chaikulngamdee

epoch	train_loss	valid_loss	accuracy	time
0	0.704212	0.707301	0.559091	01:14
1	0.582890	0.498229	0.843939	01:14
2	0.502958	0.454466	0.824242	01:18
3	0.459400	0.430813	0.853030	01:26
4	0.459093	0.417426	0.892424	01:24
5	0.414254	0.349961	0.913636	01:22
6	0.387793	0.371689	0.907576	01:22
7	0.337894	0.315346	0.940909	01:19
8	0.292890	0.277478	0.960606	01:21
9	0.259528	0.267797	0.965151	01:22

Train univariate time series

Yoga Dataset

Training Model

Graphs