Kaggle Career Con 2019

Predict surface type for robot navigation in Kaggle Career Con 2019 competition.

Prerequisites

Anaconda >=5.x

API credentials

To use the Kaggle API, sign up for a Kaggle account at https://www.kaggle.com. Then go to the 'Account' tab of your user profile (https://www.kaggle.com/<username>/account) and select 'Create API Token'. This will trigger the download of kaggle.json, a file containing your API credentials. Place this file in the location ~/.kaggle/kaggle.json (on Windows in the location C:\Users\<Windows-username>\.kaggle\kaggle.json).

For your security, ensure that other users of your computer do not have read access to your credentials. On Unix-based systems you can do this with the following command:

chmod 600 ~/.kaggle/kaggle.json

Installation guide

Set up conda environment

Using conda:

conda env create -f environment.yml
activate kaggle_career_con_2019

The packages necessary to run the project are now installed inside the conda environment.

Note: The following sections assume you are located in your conda environment.

Set up project's module

To move beyond notebook prototyping, all reusable code should go into the src/ folder package. To use that package inside your project, install the project's module in editable mode, so you can edit files in the src/ folder and use the modules inside your notebooks :

pip install --editable .

To use the module inside your notebooks, add %autoreload at the top of your notebook :

%load_ext autoreload
%autoreload 2

Example of module usage :

from src.data.make_dataset import generate
generate(10)

Set up Git diff for notebooks and lab

We use nbdime for diffing and merging Jupyter notebooks.

To configure it to this git project :

nbdime config-git --enable

To enable notebook extension :

nbdime extensions --enable --sys-prefix

Or, if you prefer full control, you can run the individual steps:

jupyter serverextension enable --py nbdime --sys-prefix

jupyter nbextension install --py nbdime --sys-prefix
jupyter nbextension enable --py nbdime --sys-prefix

jupyter labextension install nbdime-jupyterlab

You may need to rebuild the extension : jupyter lab build

Set up Plotly for Jupyterlab

Plotly works in notebook but further steps are needed for it to work in Jupyterlab :

@jupyter-widgets/jupyterlab-manager # Jupyter widgets support
plotlywidget # FigureWidget support
@jupyterlab/plotly-extension # offline iplot support

There are conflict versions between those extensions so check the latest Plotly README to ensure you fetch the correct ones.

jupyter labextension install @jupyter-widgets/jupyterlab-manager@0.36 --no-build
jupyter labextension install plotlywidget@0.2.1  --no-build
jupyter labextension install @jupyterlab/plotly-extension@0.16  --no-build
jupyter lab build

Invoke command

We use Invoke to manage an unique entry point into all of the project tasks.

List of all tasks for project :

$ invoke -l

Available tasks:

  lab     Launch Jupyter lab

Help on a particular task :

$ invoke --help lab
Usage: inv[oke] [--core-opts] notebook [--options] [other tasks here ...]

Docstring:
  Launch Jupyter lab

Options:
  -i STRING, --ip=STRING   IP to listen on, defaults to *
  -p, --port               Port to listen on, defaults to 8888

You will find the definition of each task inside the tasks.py file, so you can add your own.

PS : we don't use Makefile because some people work on Windows workstations and the install of make is cumbersome on those.

Project organization

├── tasks.py           <- Invoke with commands like `notebook`
├── README.md          <- The top-level README for developers using this project.
├── data
│   ├── external       <- Data from third party sources.
│   ├── interim        <- Intermediate data that has been transformed.
│   ├── processed      <- The final, canonical data sets for modeling.
│   └── raw            <- The original, immutable data dump.
│
├── models             <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks          <- Jupyter notebooks. Naming convention is a number (for ordering),
│                         the creator's initials, and a short `-` delimited description, e.g.
│                         `1.0-jqp-initial-data-exploration`.
│
├── references         <- Data dictionaries, manuals, and all other explanatory materials.
│
├── reports            <- Generated analysis as HTML, PDF, LaTeX, etc.
│   └── figures        <- Generated graphics and figures to be used in reporting
│
├── environment.yml    <- The requirements file for reproducing the analysis environment
│
└── src                <- Source code for use in this project.
    ├── __init__.py    <- Makes src a Python module
    │
    ├── data           <- Scripts to download or generate data
    │   └── make_dataset.py
    │
    ├── features       <- Scripts to turn raw data into features for modeling
    │   └── build_features.py
    │
    ├── models         <- Scripts to train models and then use trained models to make
    │   │                 predictions
    │   ├── predict_model.py
    │   └── train_model.py
    │
    └── visualization  <- Scripts to create exploratory and results oriented visualizations
        └── visualize.py

Project based on the cookiecutter Kaggle template project. #cookiecutterdatascience

namdnguyen / kaggle-career-con-2019