Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/tcapelle/timeseries_fastai

fastai V2 implementation of Timeseries classification papers.
https://github.com/tcapelle/timeseries_fastai

classification fastai pytorch timeseries ucr

Last synced: about 22 hours ago
JSON representation

fastai V2 implementation of Timeseries classification papers.

Host: GitHub
URL: https://github.com/tcapelle/timeseries_fastai
Owner: tcapelle
License: apache-2.0
Created: 2019-09-26T08:47:46.000Z (about 5 years ago)
Default Branch: master
Last Pushed: 2022-10-20T05:58:06.000Z (about 2 years ago)
Last Synced: 2024-10-30T21:03:43.277Z (15 days ago)
Topics: classification, fastai, pytorch, timeseries, ucr
Language: Jupyter Notebook
Homepage: https://tcapelle.github.io/timeseries_fastai/
Size: 7.77 MB
Stars: 239
Watchers: 11
Forks: 18
Open Issues: 3
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE

Awesome Lists containing this project

README

# timeseries_fastai
> This repository aims to implement TimeSeries classification/regression algorithms. It makes extensive use of fastai V2!

> I recommend to use Ignacio's [tsai](https://github.com/timeseriesAI/tsai) for a more complete and robust timeseries fastai based library. It is well documented and implemetns way more models that me here.

## Installation

You will need to install fastai V2 from [here](https://github.com/fastai/fastai) and then you can do from within the environment where you installed fastai V2:

```bash
pip install timeseries_fastai
```

and you are good to go.

### TL;DR
```bash
git clone https://github.com/fastai/fastai
cd fastai
conda env create -f environment.yml
source activate fastai
pip install fastai timeseries_fastai

```

## Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline
The original paper repo is [here](https://github.com/cauchyturing/UCR_Time_Series_Classification_Deep_Learning_Baseline) 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](https://arxiv.org/abs/1611.06455).

## InceptionTime: Finding AlexNet for Time SeriesClassification
The original paper repo is [here](https://github.com/hfawaz/InceptionTime)

- Notebook 02: Added InceptionTime architecture from [InceptionTime: Finding AlexNet for Time SeriesClassification](https://arxiv.org/pdf/1909.04939.pdf).

## Results

You can run the benchmark using:

`$python ucr.py --arch='inception' --tasks='all' --filename='inception.csv' --mixup=0.2`

### Default Values:
- `lr` = 1e-3
- `opt` = 'ranger'
- `epochs` = 40
- `fp16` = True

```python
results_inception = pd.read_csv(Path.cwd().parent/'inception.csv', index_col=0)
display_df(results_inception)
```

acc
acc_max
train_loss
val_loss

task

ACSF1
0.82
0.85
0.77
0.62

Adiac
0.77
0.77
0.81
0.89

ArrowHead
0.70
0.76
0.28
1.21

BME
0.85
0.88
0.21
0.79

Beef
0.77
0.83
0.50
0.53

BeetleFly
0.70
0.85
0.14
0.79

BirdChicken
0.95
0.95
0.14
0.20

CBF
0.95
0.97
0.22
0.24

Car
0.60
0.68
0.33
1.23

Chinatown
0.95
0.96
0.05
0.27

ChlorineConcentration
0.82
0.82
0.28
0.48

CinCECGTorso
0.58
0.60
0.42
1.60

Coffee
0.71
0.82
0.16
0.71

Computers
0.66
0.72
0.24
0.72

CricketX
0.72
0.73
0.49
0.88

CricketY
0.71
0.72
0.53
0.84

CricketZ
0.77
0.78
0.52
0.79

Crop
0.78
0.78
0.56
0.76

DiatomSizeReduction
0.93
0.96
0.22
0.22

DistalPhalanxOutlineAgeGroup
0.71
0.75
0.18
0.80

DistalPhalanxOutlineCorrect
0.74
0.78
0.16
0.57

DistalPhalanxTW
0.62
0.68
0.27
1.22

ECG200
0.87
0.91
0.15
0.30

ECG5000
0.94
0.94
0.17
0.27

ECGFiveDays
0.92
0.94
0.14
0.21

EOGHorizontalSignal
0.36
0.40
0.63
2.05

EOGVerticalSignal
0.37
0.39
0.79
2.00

Earthquakes
0.75
0.75
0.12
0.89

ElectricDevices
0.71
0.72
0.36
1.20

EthanolLevel
0.32
0.36
0.61
1.81

FaceAll
0.77
0.78
0.46
0.84

FaceFour
0.83
0.89
0.29
0.57

FacesUCR
0.83
0.83
0.51
0.73

FiftyWords
0.67
0.69
0.70
1.27

Fish
0.83
0.83
0.45
1.69

FordA
0.95
0.95
0.18
0.13

FordB
0.83
0.85
0.16
0.38

FreezerRegularTrain
0.98
0.99
0.20
0.10

FreezerSmallTrain
0.71
0.81
0.21
1.54

Fungi
0.77
0.85
0.31
0.68

GunPoint
0.95
0.97
0.17
0.14

GunPointAgeSpan
0.97
0.98
0.25
0.08

GunPointMaleVersusFemale
1.00
1.00
0.17
0.02

GunPointOldVersusYoung
1.00
1.00
0.13
0.01

Ham
0.55
0.66
0.21
1.12

HandOutlines
0.89
0.91
0.25
0.29

Haptics
0.38
0.43
0.44
1.94

Herring
0.61
0.70
0.19
0.82

HouseTwenty
0.85
0.88
0.18
0.39

InlineSkate
0.30
0.31
0.95
2.05

InsectEPGRegularTrain
1.00
1.00
0.28
0.08

InsectEPGSmallTrain
0.80
1.00
0.49
0.48

InsectWingbeatSound
0.55
0.56
0.65
1.27

ItalyPowerDemand
0.96
0.96
0.14
0.16

LargeKitchenAppliances
0.85
0.86
0.28
0.69

Lightning2
0.70
0.77
0.18
0.73

Lightning7
0.71
0.73
0.46
1.10

Mallat
0.65
0.66
0.43
1.37

Meat
0.93
0.95
0.25
0.26

MedicalImages
0.72
0.75
0.40
0.85

MelbournePedestrian
0.10
0.10
nan
nan

MiddlePhalanxOutlineAgeGroup
0.53
0.60
0.20
1.28

MiddlePhalanxOutlineCorrect
0.77
0.81
0.17
0.46

MiddlePhalanxTW
0.49
0.59
0.34
1.37

MixedShapesRegularTrain
0.93
0.93
0.35
0.25

MixedShapesSmallTrain
0.80
0.81
0.42
0.64

MoteStrain
0.75
0.76
0.09
0.52

NonInvasiveFetalECGThorax1
0.92
0.93
0.66
0.32

NonInvasiveFetalECGThorax2
0.93
0.93
0.59
0.27

OSULeaf
0.82
0.84
0.43
0.58

OliveOil
0.77
0.80
0.27
0.74

PhalangesOutlinesCorrect
0.81
0.83
0.17
0.46

Phoneme
0.22
0.22
0.79
3.25

PigAirwayPressure
0.12
0.14
2.33
4.06

PigArtPressure
0.47
0.47
1.25
2.25

PigCVP
0.30
0.33
1.69
2.97

Plane
1.00
1.00
0.35
0.07

PowerCons
0.98
0.98
0.17
0.10

ProximalPhalanxOutlineAgeGroup
0.83
0.87
0.22
0.53

ProximalPhalanxOutlineCorrect
0.88
0.89
0.17
0.34

ProximalPhalanxTW
0.78
0.80
0.28
0.78

RefrigerationDevices
0.50
0.56
0.27
1.35

Rock
0.58
0.78
0.29
1.43

ScreenType
0.42
0.43
0.33
1.41

SemgHandGenderCh2
0.73
0.79
0.21
0.52

SemgHandMovementCh2
0.35
0.40
0.43
1.56

SemgHandSubjectCh2
0.52
0.52
0.39
1.13

ShapeletSim
0.99
1.00
0.14
0.12

ShapesAll
0.80
0.80
0.89
0.83

SmallKitchenAppliances
0.65
0.66
0.43
1.60

SmoothSubspace
0.96
0.97
0.23
0.15

SonyAIBORobotSurface1
0.87
0.90
0.08
0.29

SonyAIBORobotSurface2
0.75
0.79
0.15
0.54

StarLightCurves
0.98
0.98
0.22
0.09

Strawberry
0.97
0.98
0.15
0.09

SwedishLeaf
0.94
0.94
0.52
0.27

Symbols
0.83
0.87
0.39
0.61

SyntheticControl
1.00
1.00
0.31
0.04

ToeSegmentation1
0.93
0.97
0.16
0.17

ToeSegmentation2
0.88
0.91
0.15
0.27

Trace
1.00
1.00
0.29
0.02

TwoLeadECG
0.91
0.92
0.10
0.26

TwoPatterns
1.00
1.00
0.25
0.01

UMD
0.92
0.94
0.25
0.26

UWaveGestureLibraryAll
0.91
0.91
0.41
0.31

UWaveGestureLibraryX
0.82
0.82
0.46
0.56

UWaveGestureLibraryY
0.73
0.73
0.50
0.78

UWaveGestureLibraryZ
0.74
0.74
0.48
0.72

Wafer
1.00
1.00
0.05
0.01

Wine
0.48
0.63
0.19
1.07

WordSynonyms
0.62
0.62
0.61
1.60

Worms
0.77
0.78
0.41
0.70

WormsTwoClass
0.73
0.81
0.22
0.56

Yoga
0.86
0.86
0.24
0.33

## Getting Started

```python
from timeseries_fastai.imports import *
from timeseries_fastai.core import *
from timeseries_fastai.data import *
from timeseries_fastai.models import *
```

```python
PATH = Path.cwd().parent
```

```python
df_train, df_test = load_df_ucr(PATH, 'Adiac')
```

Loading files from: /home/tcapelle/SteadySun/timeseries_fastai/Adiac

```python
x_cols = df_train.columns[0:-2].to_list()
```

```python
dls = TSDataLoaders.from_dfs(df_train, df_test, x_cols=x_cols, label_col='target', bs=16)
```

```python
dls.show_batch()
```

![png](docs/images/output_17_0.png)

```python
inception = create_inception(1, len(dls.vocab))
```

```python
learn = Learner(dls, inception, metrics=[accuracy])
```

```python
learn.fit_one_cycle(1, 1e-3)
```

epoch train_loss valid_loss accuracy time
0 3.934007 3.640701 0.043478 00:03