Dive Into Learning From Data

Описание: Dive into Learning From Data using the famous MNIST dataset. Explore the classical Logistic Regression algorithm, a simple yet powerful method, and aim to achieve an almost perfect score on MNIST. We review the fundamentals, analyze the data, and evaluate performance metrics. Finally, we apply dimensionality reduction (PCA) and feature engineering (Polynomial Features) to boost performance, achieving an impressive 98.05% accuracy.

Check my blog post for the text content:
https://datasanta.net/2024/11/26/dive...

The Jupyter notebook can be found here:
https://github.com/nickovchinnikov/da...

Join me on:
Telegram: https://t.me/datasantaa
Twitter: https://x.com/datasantaa

0:07 - Intro to Learning from Data
1:15 - Understanding MNIST Dataset
3:00 - Data Analysis Overview
5:04 - Visualizing MNIST Images
7:16 - What is Logistic Regression?
7:39 - Explaining the Sigmoid Function
11:09 - Mechanics of Logistic Regression
12:20 - Simulating Logistic Regression
16:30 - Implementing Logistic Regression with sklearn
17:45 - Splitting Data into Test and Train Sets
20:40 - Initial Training Results
21:45 - Evaluating Model Performance
25:35 - Introduction to Confusion Matrix
29:17 - Constructing the Confusion Matrix
30:34 - Analyzing MNIST Confusion Matrix
34:35 - Calculating Accuracy from Confusion Matrix
36:24 - Understanding Precision
46:25 - Understanding Recall
49:45 - F1-Score: Balancing Precision and Recall
53:24 - Comparing All Key Metrics at Once
57:07 - Data Preprocessing and Feature Engineering Techniques
58:24 - StandardScaler
1:05:46 - MinMaxScaler
1:07:59 - Principal Component Analysis (PCA)
1:19:35 - Polynomial Features
1:24:18 - Results