certifaiai / time-series-labs Goto Github PK

View Code? Open in Web Editor NEW

18.0 1.0 13.0 118.11 MB

Hands-on training labs for Deep Learning in Time Series course by CERTIFAI

Home Page: https://certifai.ai/

License: GNU Affero General Public License v3.0

Jupyter Notebook 99.92% Python 0.08%

deep-learning time-series

time-series-labs's Introduction

Time Series Labs

This repository contains Jupyter notebooks used for training during time series course delivery.

AR Modelling
ARIMA
Multilayer perceptron in time series forecasting
Time series forecasting using deep learning
Time series classification and anomaly detection using deep learning

Getting Started

Install Anaconda Individual Edition

Download and install Anaconda.

Environment Setup

You can install the dependencies through any of the following ways:

Setup the virtual environment using conda by

conda env create -f environment.yml

Setup the virtual environment using virtualenv with Python version 3.8 by

pip install -r requirement.txt

The environment setup will take some time to download required modules.

GPU Setup (optional)

Follow the instructions below if you plan to use GPU setup.

Install CUDA and cuDNN Requirements:
- CUDA 10.2
- cuDNN 7.6

Step by step installation guides can be found here.

If you like to use different version of CUDA, please install appropriate cudatoolkit module by enter conda install cudatoolkit=CUDA_VERSION

conda install cudatoolkit=10.2

Usage

All examples are separated into [training] and [solution] folders.

All notebooks in training folder have few lines commented out so that they can be taught and demonstrated in the class. The solution folder contains the un-commented version for every line of codes.

Known Issues

time-series-labs's People

Contributors

Stargazers

Watchers

Forkers

afiqmuzaffar praneshkrishnan kurucan fareezjusman-skymind amirunpri2018 alokproc tingweijing evanyeohboonkhai yuenherny hazirahindra code-kaki dimiphoton

time-series-labs's Issues

03 - Data Preprocessing Techniques typo

Eg. 2.2.3 - Cap the outliers with a certain range
clapped_series => capped_series

04 - CNN_Multivariate_Univariate_SingleStep normalization issue

For multivariate input data, each column should perform normalization differently instead of reshape into one column and perform scaling.

The correct way should be similar to image below:

03 - Data Preprocessing Techniques

Describe the bug
Eg. 2.2.2 - Correct them with linear interpolation

To Reproduce
Steps to reproduce the behavior:
Run the notebooks

Expected behavior
It should replace with NaN value and perform interpolatiob.
It is not replaced with a zero value .

** Screenshot **

Syntax error for `testPC.py` in `solution/` folder

Describe the bug
Syntax error for testPC.py in solution/ folder

To Reproduce
Steps to reproduce the behavior:

Go to testPC.py and execute the file.

Expected behavior
Executable file.

Screenshots

gist-it unavailable on JupyterLab

Suggested Edits

Users reported gist-it was not available in JupyterLab. Suggest to check on ways to configure it or provide alternative (eg. using Jupyter Notebook) in WIki or in installation tutorial.

Syntax Error for notebook `04 - Hybrid_CNN_LSTM_Univariate.ipynb` and `04 - CNN_Univariate_Multistep.ipynb`

Describe the bug
Syntax error for code cells below for notebook 04 - Hybrid_CNN_LSTM_Univariate.ipynb and 04 - CNN_Univariate_Multistep.ipynb

To Reproduce
Steps to reproduce the behavior:

Go to the below mentioned notebooks and check the cell as shown in the image. Try to execute it and it will result in syntax error.

Expected behavior
Executable cell.

Screenshots

Conda package cannot be resolved for macOS

Describe the bug

Package cudatoolkit cannot be resolved while trying to install the conda environment for macOS.

To Reproduce
Steps to reproduce the behavior:

Tried to create the environment based on the command conda env create -f environment.yml.
Error triggered as shown.

Expected behavior

Conda environment can be created.

Screenshots

After did some checking at the Anaconda registry, identified that the cudatoolkit package only support macOS only up to version 9.0

Suggest to remove the set version for cudatoolkit package in the environment file so that it automatically download & use the compatible version for respective OSes.

Desktop (please complete the following information):

OS: macOS
Version: 10.15.7 (19H2)

"Untitled" Notebook

Found a "Untitled" notebook in the repo. Suggest to remove it as it looks like it is a notebook that was developed earlier but renamed to other title, without deleting the notebook.

Attached notebook link here: https://github.com/CertifaiAI/time-series-labs/blob/main/notebook/Solution/Untitled.ipynb

test

Include requirments.txt for list of packages when install using virtualenv

Additional context

Some users might choose to not using conda to follow along with this lab. Suggest to have a requirement.txt for users planning to follow along using virtualenv.

03 - Data Preprocessing Techniques

Eg. 2.2.1 should be fill with np.nan instead of "NaN"
this does not work with pd.Series().fillna() method

Incorrect variable naming in instruction

Describe the bug
Incorrect variable name in instruction at practice part for Simple Exponential Moving Average.

To Reproduce
Steps to reproduce the behavior:

Go to 'notebook/Training/01 - Introduction to Forecasting Methods - Moving Averages and Holt's Methods.ipynb'
Scroll down to practice part for Simple Exponential Moving Average.

Expected behavior
'fit_1' should be 'fit1'
'fit_2' should be 'fit2'
'fit_3' should be 'fit3'

Screenshots

inverse transformation instructions can be clearer

Current inverse transformation code on Day 2 ARIMA is quite messy. Revise for better understanding.

certifaiai / time-series-labs Goto Github PK

time-series-labs's Introduction

Time Series Labs

Contents

Getting Started

Install Anaconda Individual Edition

Environment Setup

GPU Setup (optional)

Usage

Known Issues

time-series-labs's People

Contributors

Stargazers

Watchers

Forkers

time-series-labs's Issues

Suggested Edits

Additional context

Recommend Projects

Recommend Topics

Recommend Org