Dive Into Learning From Data: MNIST with Logistic Regression

Name: Dive Into Learning From Data: MNIST with Logistic Regression
Rating: 4.2 (2 reviews)

Master Classification with Python: Learn logistic regression, PCA, and feature engineering to achieve 98% accuracy!

Created byNick Ovchinnikov

Last updated 2/2025

English

English [Auto],

What you'll learn

Understand the Fundamentals of Classification: how to work with the MNIST dataset, preprocess image data, and visualize handwritten digits
Build and Train a Logistic Regression Classifier from scratch, understand the mathematics behind it (e.g., sigmoid function)
Evaluate Model Performance: accuracy, precision, recall, F1 score, and confusion matrices
Apply Advanced Techniques for Better Performance: Principal Component Analysis (PCA) for dimensionality reduction and Polynomial Feature Expansion

Course content

3 sections • 12 lectures • 1h 50m total length

Understanding MNIST Dataset6:09
Explore the MNIST dataset by loading data and target vectors, inspecting 784-pixel images, and visualizing centered digits with matplotlib to understand how grayscale pixel intensity reveals digits.
What is Logistic Regression?5:10
Explore how logistic regression classifies digits by applying a sigmoid to a linear combination of weights and bias, converting to a probability, and using a 0.5 threshold for class labels.
Simulating Logistic Regression5:18
Simulate logistic regression with 100 samples and 3 features, generate random weights and bias, compute z = x dot weights plus bias, apply sigmoid to get y_hat, and plot results.
Matrix Multiplication20:12
Explore how matrix multiplication maps input x to weights w and bias b, using dot products, dimensionality rules, and broadcasting in deep learning.
The Broadcasting rules5:51
Learn how the broadcasting rules enable operations on arrays of different shapes, including stretching dimensions, aligning shapes, and performing addition and matrix multiplication with concrete examples.

Splitting Data into Test and Train Sets3:52
Split data into train and test (80/20), train a logistic regression model, predict on the test set, and evaluate with accuracy, noting convergence may require more iterations or data scaling.
Constructing the Confusion Matrix12:57
Construct a ten-by-ten MNIST confusion matrix to visualize misclassifications, identify true positives, false positives, and false negatives, and compute accuracy, precision, recall, and F1.
Calculating Accuracy from Confusion Matrix15:10
Compute accuracy from the confusion matrix as the trace divided by the total sum, and examine per-class precision and recall using true positives and false positives.
Understanding Precision / Recall7:13
Explore how the F1 score, the harmonic mean of precision and recall, provides a balanced metric for MNIST classification, with per-class pandas visualizations and logistic regression with balanced weights.

Feature Engineering Techniques11:00
Improve MNIST digit classification by standardizing data with a standard scaler and using a min-max scaler to assess logistic regression performance.
Principal Component Analysis (PCA)11:46
Use principal component analysis to reduce high-dimensional data by capturing directions with the highest variance, transforming training data, and reconstructing images for accelerated logistic regression with PCA features.
Polynomial Features and Results5:45
Reduce to two PCA components, then apply degree-two polynomial features to create squared and interaction terms; train logistic regression on the transformed data to achieve a 98% score.

Requirements

Familiarity with Python programming is recommended. You should be comfortable with basic syntax, data structures (e.g., lists, dictionaries), and control flow (e.g., loops, conditionals).
A basic understanding of linear algebra (e.g., vectors, matrices) and probability will help you grasp concepts like Logistic Regression and PCA. Don’t worry—we’ll break everything down step by step!
You'll need a Python environment set up with libraries like NumPy, pandas, matplotlib, and scikit-learn
No prior experience in machine learning is required! This course is designed to be beginner-friendly, with clear explanations and hands-on examples to guide you every step of the way.

Description

Unlock the Power of Image Classification with Python!

Are you ready to dive into the fascinating world of image classification? In this comprehensive course, you'll learn how to teach a computer to recognize and classify images using Python. Whether you're a beginner or an experienced data scientist, this course will guide you through the entire process of building, training, and evaluating image classification models.

Handwritten Digit Recognition — Learn Everything You Need to Start Your Machine Learning Journey in One Comprehensive Course!

What You'll Learn:

Introduction to Image Classification: Understand the fundamentals of image classification and explore the MNIST dataset, a collection of handwritten digits.
Data Preprocessing: Learn how to preprocess and visualize image data using Python libraries like matplotlib and scikit-learn.
Building a Simple Classifier: Implement a logistic regression model to classify handwritten digits and understand the underlying mathematics, including the sigmoid function.
Model Evaluation: Dive into model evaluation techniques, including accuracy, precision, recall, and F1 score. Learn how to interpret confusion matrices and improve model performance.
Advanced Techniques: Explore advanced techniques like Principal Component Analysis (PCA) for dimensionality reduction and polynomial feature expansion to capture complex relationships in the data.
Optimization: Discover how to fine-tune your models by scaling data, balancing class weights, and optimizing hyperparameters.

Prerequisites:

Basic knowledge of Python programming.
Familiarity with basic machine learning concepts (helpful but not required).

Who Is This Course For?

Aspiring data scientists and machine learning enthusiasts who want to learn image classification from scratch.
Python developers looking to expand their skill set into machine learning and computer vision.
Professionals who want to understand the theory and practical implementation of image classification models.

By the End of This Course, You'll Be Able To:

Preprocess and visualize image data effectively.
Build and train image classification models using logistic regression.
Evaluate and interpret model performance using various metrics.
Apply advanced techniques like PCA and polynomial feature expansion to improve model accuracy.
Fine-tune models for optimal performance.

Enroll Now and Start Your Journey into Image Classification with Python!

Who this course is for:

Data Science Beginners & Python Programmers
Software Developers Transitioning to ML

Dive Into Learning From Data: MNIST with Logistic Regression

What you'll learn

Explore related topics

Course content

Introduction5 lectures • 43min

Training4 lectures • 39min

Feature Engineering3 lectures • 29min

Requirements

Description

Who this course is for: