Machine Learning Experimentation solving MNIST

---

Objective

In this notebook, we will learn to experiment solving MNIST dataset using several different learning algorithm e.g. Logistic Regression, Random Forest, Multi Layer Perceptron (MLP), and Convolutional Neural Network (CNN) while tracking the experiment using MLfow

Problem

Given 28x28 pixels of handwritten digit, predict the number (0-9), hence this is multi-class classification problem.

Evaluation Metrics

• Accuracy
• F1 score
• Precision
• Recall
• Time taken for training
• Time taken for predicting

---

Environment

This notebook run using python 3.7. I use Anaconda to set the environment.

conda create --name new_environment python=3.7 # create new environment
conda activate new_environment # activate the env
pip install -r requirements.txt # install packages in requirements.txt