Skip to content

Latest commit

 

History

History
80 lines (67 loc) · 2.53 KB

README.md

File metadata and controls

80 lines (67 loc) · 2.53 KB

KaggleZeroToAll

After knowing basics of machine learning, deep learning, and TensorFlow/Keras, what's the next?

Kaggle provides many interesting problems for machine learning experts. This repository hosts interesting Kaggle problems and show how to solve these problems using decent deep learning models.

Kaggle problem directory naming

k0-00-short-title

  • Difficulty (k0, k1, ... k9):
    • 0: easy
    • 5: normal
    • 9: very difficult
  • k0-XX: 00 serial number
  • short-title: title for the Kaggle data
  • put .py, .ipynb, and data files in the directory
    • If data files are large, you can create a script. Please check this

Content of each file

Please see k0-00-template.ipynb

  • Kaggle name
  • dataset/problem description
  • loading data
  • model to solve the problem
  • results
  • future work and exercises

Dependencies for Kaggle Utils (optional)

requests==2.13.0
beautifulsoup4==4.6.0

or

pip install -r requirements.txt

Kaggle Utils (optional)

  • kaggle_download.py: Kaggle download script

    1. Create kaggle.ini
      • Copy kaggle.ini.sample and name it kaggle.ini
      • Fill out your username and password in kaggle.ini
    2. Accept the agreement term in Kaggle website
      • Click the download button on the competition main site
    3. Find a competition name
    4. In terminal,
    # python kaggle_download.py competition-name --destination path/to/save/dataset
    # Example:
    $ python kaggle_download.py digit-recognizer --destination k0-01-mnist/input
  • kaggle_submit.py: Kaggle submission script

    1. You can also submit your submission
    2. In terminal,
    # python kaggle_submit.py competition-name /path/to/submission.csv -m "Submission message"
    # Example:
    python kaggle_submit.py digit-recognizer k0-01-mnist/submission.csv -m "First Submission"

Tests

py.test

Contributions

We welcome any contributions including writing issues and sending pull requests.

References (Thanks to the TF-KR user group)