TimeSeries_fastai

This repository aims to implement TimeSeries classification/regression algorithms. It makes extensive use of fastai training methods.

Installation

In short, if you have anaconda, execute:

$ conda env create --file=environment.yml
$ conda activate timseries
$ jupyter notebook

Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline

The original paper repo is here is implemented in Keras/Tf.

Notebook 01: This is a basic notebook that implements the Deep Learning models proposed in Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline.

InceptionTime: Finding AlexNet for Time SeriesClassification

The original paper repo is here

Notebook 01_inception: Added InceptionTime architecture from InceptionTime: Finding AlexNet for Time SeriesClassification. Need to check if the implementation is identical to paper, please comment.

The model is implemented in inception.py, you can also find an inception-resnet implemenation.

The ucr.py script can be used interactively. You can display all possible options with:

$ python ucr.py --help

usage: ucr.py [-h] [--arch ARCH] [--tasks TASKS] [--epochs EPOCHS] [--lr LR]
Training UCR script

optional arguments:
  -h, --help       show this help message and exit
  --arch ARCH      Network arch. [resnet, FCN, MLP, iresnet, inception, All]. (default: 'resnet')
  --tasks TASKS    Which tasks from UCR to run, [task, All]. (default: 'Adiac')
  --epochs EPOCHS  Number of epochs.(default: 40)
  --lr LR          Learning rate.(default: 1e-3)

To run resnet in the Adiac data set:

$ python ucr.py

The default values are 40 epochs and lr=1e-3. You can modify this using the epochs and lr arguments when calling ucr.

$ python ucr.py --epochs=100 --lr=1e-3

You can also chosse the architecture and the task to run with params arch and tasks.

$ python ucr.py --epochs=30 --lr=1e-3 --tasks='Adiac' --arch='fcn'

To run everything with default settings:

$ python ucr.py --tasks='All' --arch='All'

The whole dataset runs in less than one hour on a RTX2080ti with default settings.

Results

Params: epochs = 40 and lr=1e-3.

from tabulate import tabulate
print(tabulate(results,  tablefmt="pipe", headers=results.columns))

	MLP	FCN	resnet	iresnet	inception
ACSF1	0.33	0.19	0.17	0.16	0.13
Adiac	0.29156	0.237852	0.214834	0.225064	0.253197
AllGestureWiimoteX	0.9	0.9	0.9	0.9	0.9
AllGestureWiimoteY	0.9	0.9	0.9	0.9	0.9
AllGestureWiimoteZ	0.9	0.9	0.9	0.9	0.9
ArrowHead	0.228571	0.257143	0.222857	0.262857	0.12
BME	0.0333333	0	0.00666667	0.24	0
Beef	0.233333	0.266667	0.233333	0.233333	0.3
BeetleFly	0.2	0.1	0.15	0.1	0
BirdChicken	0.35	0.05	0.05	0.1	0
CBF	0.115556	0	0.00444444	0.00333333	0
Car	0.166667	0.166667	0.166667	0.166667	0.0833333
Chinatown	0.0233236	0.00874636	0.0145773	0.0174927	0.0116959
ChlorineConcentration	0.304427	0.278906	0.23151	0.278646	0.16276
CinCECGTorso	0.268116	0.226812	0.210145	0.269565	0.208696
Coffee	0	0	0	0	0
Computers	0.388	0.24	0.204	0.188	0.184
CricketX	0.420513	0.246154	0.217949	0.228205	0.179487
CricketY	0.397436	0.24359	0.225641	0.269231	0.184615
CricketZ	0.417949	0.179487	0.176923	0.212821	0.14359
Crop	0.322143	0.22875	0.225298	0.229821	0.222619
DiatomSizeReduction	0.0751634	0.114379	0.104575	0.0816993	0.0522876
DistalPhalanxOutlineAgeGroup	0.244604	0.223022	0.230216	0.223022	0.223022
DistalPhalanxOutlineCorrect	0.26087	0.199275	0.199275	0.199275	0.199275
DistalPhalanxTW	0.294964	0.294964	0.280576	0.266187	0.244604
DodgerLoopDay	0.85	0.85	0.85	0.85	0.85
DodgerLoopGame	0.478261	0.478261	0.478261	0.478261	0.478261
DodgerLoopWeekend	0.26087	0.26087	0.26087	0.26087	0.26087
ECG200	0.09	0.06	0.07	0.08	0.08
ECG5000	0.0557778	0.0528889	0.0548889	0.0562222	0.0511111
ECGFiveDays	0.0732558	0.0244186	0.0255814	0.0860465	0
EOGHorizontalSignal	0.51105	0.348066	0.345304	0.359116	0.359116
EOGVerticalSignal	0.571823	0.516575	0.524862	0.541436	0.48895
Earthquakes	0.258993	0.230216	0.251799	0.230216	0.244604
ElectricDevices	0.426015	0.277655	0.267151	0.241084	0.258332
EthanolLevel	0.314	0.226	0.236	0.366	0.146
FaceAll	0.25858	0.0662722	0.0508876	0.0556213	0.176331
FaceFour	0.113636	0.113636	0.102273	0.159091	0.0454545
FacesUCR	0.198049	0.0570732	0.0560976	0.0809756	0.037561
FiftyWords	0.265934	0.384615	0.362637	0.421978	0.268132
Fish	0.114286	0.0228571	0.0685714	0.0571429	0.0114286
FordA	0.155303	0.0636364	0.0590909	0.0636364	0.0356061
FordB	0.279012	0.17037	0.179012	0.2	0.144444
FreezerRegularTrain	0.0210526	0.00140351	0.00280702	0.00245614	0.00140351
FreezerSmallTrain	0.225263	0.0119298	0.0119298	0.0333333	0.0968421
Fungi	0.505376	0.306452	0.198925	0.483871	0.241935
GestureMidAirD1	0.961538	0.961538	0.961538	0.961538	0.961538
GestureMidAirD2	0.961538	0.961538	0.961538	0.961538	0.961538
GestureMidAirD3	0.961538	0.961538	0.961538	0.961538	0.961538
GesturePebbleZ1	0.837209	0.837209	0.837209	0.837209	0.837209
GesturePebbleZ2	0.848101	0.848101	0.848101	0.848101	0.848101
GunPoint	0.0333333	0	0	0	0
GunPointAgeSpan	0.0411392	0.028481	0.028481	0.0253165	0.0158228
GunPointMaleVersusFemale	0.0126582	0	0	0	0
GunPointOldVersusYoung	0	0	0	0	0
Ham	0.219048	0.161905	0.2	0.228571	0.2
HandOutlines	0.072973	0.0648649	0.0567568	0.183784	0.0432432
Haptics	0.50974	0.516234	0.512987	0.558442	0.451299
Herring	0.21875	0.296875	0.25	0.34375	0.28125
HouseTwenty	0.218487	0.0420168	0.0420168	0.0504202	0.0252101
InlineSkate	0.652727	0.578182	0.547273	0.494545	0.625455
InsectEPGRegularTrain	0	0	0	0	0
InsectEPGSmallTrain	0	0	0.475806	0.475806	0.165323
InsectWingbeatSound	0.360606	0.531818	0.515152	0.567677	0.361111
ItalyPowerDemand	0.0330418	0.0281827	0.0281827	0.0349854	0.0252672
LargeKitchenAppliances	0.557333	0.0906667	0.101333	0.0853333	0.101333
Lightning2	0.213115	0.180328	0.163934	0.196721	0.180328
Lightning7	0.287671	0.150685	0.123288	0.136986	0.138889
Mallat	0.155224	0.034968	0.0345416	0.0307036	0.0465017
Meat	0.0166667	0.0166667	0.0166667	0.05	0.1
MedicalImages	0.260526	0.215789	0.213158	0.227632	0.213158
MelbournePedestrian	0.899549	0.899549	0.899549	0.899549	0.899549
MiddlePhalanxOutlineAgeGroup	0.376623	0.357143	0.344156	0.357143	0.357143
MiddlePhalanxOutlineCorrect	0.158076	0.158076	0.168385	0.178694	0.171821
MiddlePhalanxTW	0.383117	0.435065	0.409091	0.415584	0.38961
MixedShapesRegularTrain	0.0779381	0.0457732	0.0507216	0.0420619	0.0305155
MixedShapesSmallTrain	0.167835	0.111753	0.107216	0.123711	0.0960825
MoteStrain	0.155751	0.116613	0.108626	0.119808	0.0926518
NonInvasiveFetalECGThorax1	0.0651399	0.0503817	0.0412214	0.0554707	nan
NonInvasiveFetalECGThorax2	0.0554707	0.0554707	0.0503817	0.0534351	nan
OSULeaf	0.429752	0.0247934	0.0454545	0.0454545	0.0495868
OliveOil	0.0666667	0.2	0.166667	0.233333	0.6
PLAID	0.938547	0.938547	0.938547	0.938547	0.938547
PhalangesOutlinesCorrect	0.188811	0.168998	0.170163	0.159674	0.162005
Phoneme	0.894515	0.686709	0.679852	0.691456	0.658755
PickupGestureWiimoteZ	0.9	0.9	0.9	0.9	0.9
PigAirwayPressure	0.870192	0.75	0.778846	0.730769	0.673077
PigArtPressure	0.764423	0.456731	0.490385	0.490385	0.0817308
PigCVP	0.875	0.8125	0.764423	0.759615	0.586538
Plane	0.0190476	0	0	0	0
PowerCons	0	0.0444444	0.0555556	0.0611111	0
ProximalPhalanxOutlineAgeGroup	0.136585	0.131707	0.126829	0.121951	0.131707
ProximalPhalanxOutlineCorrect	0.123711	0.0790378	0.0790378	0.0962199	0.0756014
ProximalPhalanxTW	0.195122	0.190244	0.17561	0.204878	0.2
RefrigerationDevices	0.592	0.429333	0.424	0.408	0.448
Rock	0.42	0.58	0.34	0.68	0.42
ScreenType	0.549333	0.421333	0.413333	0.368	0.402667
SemgHandGenderCh2	0.085	0.133333	0.131667	0.121667	0.108333
SemgHandMovementCh2	0.36	0.393333	0.397778	0.391111	0.435556
SemgHandSubjectCh2	0.0911111	0.317778	0.306667	0.302222	0.326667
ShakeGestureWiimoteZ	0.9	0.9	0.9	0.9	0.9
ShapeletSim	0.433333	0.0277778	0	0.0166667	0
ShapesAll	0.218333	0.113333	0.101667	0.136667	0.101667
SmallKitchenAppliances	0.584	0.213333	0.226667	0.197333	0.208
SmoothSubspace	0.0333333	0	0.00666667	0	0
SonyAIBORobotSurface1	0.168053	0.0332779	0.0299501	0.0349418	0.0183333
SonyAIBORobotSurface2	0.137461	0.0535152	0.0524659	0.0304302	0.0409664
StarLightCurves	0.0457746	0.0251336	0.0246479	0.0262263	0.021491
Strawberry	0.0324324	0.0297297	0.027027	0.027027	0.0297297
SwedishLeaf	0.088	0.0368	0.0384	0.0368	0.032
Symbols	0.173387	0.0544355	0.0745968	0.0725806	0.0181452
SyntheticControl	0.02	0	0.00333333	0.01	0
ToeSegmentation1	0.390351	0.0394737	0.0350877	0.0438596	0.0350877
ToeSegmentation2	0.207692	0.0692308	0.0846154	0.0846154	0.0230769
Trace	0.18	0	0	0	0
TwoLeadECG	0.108963	0.013181	0.000878735	0	0.000878735
TwoPatterns	0.078	0.03175	0.00825	0.00775	0
UMD	0.125	0.00694444	0	0.0138889	0.00694444
UWaveGestureLibraryAll	0.0457845	0.140145	0.139866	0.158012	0.0563931
UWaveGestureLibraryX	0.228643	0.217476	0.206868	0.220547	0.170854
UWaveGestureLibraryY	0.30402	0.309883	0.300391	0.329704	0.228922
UWaveGestureLibraryZ	0.28727	0.262982	0.251535	0.266611	0.226131
Wafer	0.0045425	0.00194679	0.00243348	0.00210902	0.00178456
Wine	0.166667	0.166667	0.125	0.145833	0.333333
WordSynonyms	0.374608	0.413793	0.407524	0.468652	0.327586
Worms	0.467532	0.194805	0.233766	0.25974	0.168831
WormsTwoClass	0.324675	0.194805	0.155844	0.142857	0.142857
Yoga	0.169333	0.165	0.154667	0.138	0.110667

selvakumar-sss / timeseries_fastai Goto Github PK