# Run this cell to install the latest version of fastai2 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 *

Training and End-to-End Model

This is a code-dense example. For a more detailed example showing all options found in this timeseries package please refer to this [notebook] (https: //github.com/ai-fast-track/timeseries/blob/master/nbs/index.ipynb)

Default settings:

Param	Value
model	inception_time
opt_func	Ranger
loss_func	LabelSmoothingCrossEntropy()
Normalize	per_sample_per_channel

Training a model

# You can choose any multivariate or univariate dataset listed the `data.py` file
path = unzip_data(URLs_TS.NATOPS)
fnames = [path/'NATOPS_TRAIN.arff', path/'NATOPS_TEST.arff']
dls = TSDataLoaders.from_files(bs=32,fnames=fnames, batch_tfms=[Normalize()], num_workers=0) # num_workers=0 is for Windows platform
learn = ts_learner(dls)
learn.fit_one_cycle(25, lr_max=1e-3) 

epoch	train_loss	valid_loss	accuracy	time
0	2.803271	1.788028	0.138889	00:02
1	2.640071	1.800265	0.152778	00:02
2	2.337025	1.809550	0.152778	00:02
3	1.958659	1.631379	0.375000	00:02
4	1.638520	0.996783	0.763889	00:02
5	1.393388	0.791485	0.833333	00:02
6	1.217224	0.741780	0.861111	00:02
7	1.084642	0.672229	0.861111	00:02
8	0.981334	0.743353	0.833333	00:02
9	0.901527	0.734145	0.847222	00:02
10	0.835500	0.730749	0.819444	00:02
11	0.783602	0.624519	0.888889	00:02
12	0.731175	0.746540	0.847222	00:02
13	0.688649	0.614995	0.916667	00:02
14	0.649237	0.624472	0.930556	00:02
15	0.615866	0.593726	0.944444	00:02
16	0.586982	0.589635	0.958333	00:02
17	0.562923	0.549071	0.986111	00:02
18	0.541431	0.557641	0.958333	00:02
19	0.523363	0.556317	0.944444	00:02
20	0.509103	0.574100	0.944444	00:02
21	0.495959	0.559972	0.958333	00:02
22	0.486048	0.553628	0.958333	00:02
23	0.477561	0.550643	0.958333	00:02
24	0.470691	0.549953	0.958333	00:02

Graphs

dls.show_batch(chs=range(0,12,3))

learn.recorder.plot_loss()

learn.show_results(chs=range(0,24,3))

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

interp.most_confused()

[('2.0', '1.0', 1), ('2.0', '3.0', 1), ('3.0', '2.0', 1)]