Skip to content
Case Study
Jupyter Notebook
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Freesound General-Purpose Audio Tagging Challenge
MNIST
PhysioNet 2016 Challenge
The Chinook Database
VOICED Database
README.md

README.md

Small Case Studies

A list of small case studies to educate myself in data science (databases), machine learning and deep learning.


Data Science

  • VOICED Database

    The VOICED (VOIce ICar fEDerico II) database is a database that includes 208 voice samples, from 150 pathological, and 58 healthy voices. In the notebooks the voice recordings in WFDB format (.dat) and in text format (.txt) are dowloaded together with the info text files (-info.txt) and the header files (.hea) to create a dataset saved in an excel file.

    • Download the dataset: notebook
    • Create the dataset: notebook, dataset is stored in dataset_InfoTxtFile.xlsx
  • The Chinook Database (SQL)

    In this notebook we play around with an example database using SQL and Python.


Machine Learning

  • Freesound General-Purpose Audio Tagging Challenge

    In this notebook, Mel Frequency Ceptral Coefficient (MFCC) features from the audio signal are extracte, using libROSA, and a classification using Random Forest is performed. In this notebook only two labels are selected from the dataset: bark (dog) and meow (cat).

  • PhysioNet 2016 Challenge.

    Classification of Normal/Abnormal Hearth Sounds Recordings.

    Starting from a feature table, a machine learning algorithm is explored. The feature table comes from the Heart Sound Classification Matlab demo, extracted from the PhysioNet/CinC Challenge 2016 challenge dataset.

    • Logistic regression: notebook
    • Random forest and comparison with logistic regression: notebook

Deep Learning

  • MNIST

    MNIST the "Hello World" of deep learning. In this notebook grayscale images of handwritten digits (image size: 28 x 28 pixels) are classsified into 10 categories (0 through 9), in Keras using dense layers and 2D convnets.


References

  • Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems, Aurélien Géron
  • Python Machine Learning: Machine Learning and Deep Learning with Python, scikit-learn, and TensorFlow, Sebastian Raschka and Vahid Mirjalili
  • Deep Learning with Python, Francois Chollet


Last update: 22 September 2019

You can’t perform that action at this time.