Intro

Part 1: AI, ML and Deep Learning

• AI: get machine to fill task that requires human level intelligence
• ML: use specialized algorithm by learning from data -> find patterns
• Deep Learning: rely on neural network to learn from data

This course cover the classical machine learning, includes

• ML core concepts
• statistical techs including regression, classification, clustering and more

Part II: ML history

We do not want to amplify human bias

Part III: Techniques for Machine Learning

• Decide if AI is the right approach of your problem:
  • the problem can't be defined by precise rules &
  • you can get extensive data that contains the solution
• From preparation to production

1. Data Collection and Preparation

   • First, you gather relevant data for your problem
   • Clean the data by handling missing values, outliers, and inconsistencies
   • Split the data into two or three sets:
     • Training set (typically 70-80% of data)
     • Validation set (optional, 10-15%)
     • Test set (20-30%)

2. Feature Engineering

   • Select relevant features (variables) that will help predict your target
   • Transform features through:
     • Scaling (normalizing numbers to similar ranges)
     • Encoding categorical variables
     • Creating new features from existing ones

3. Model Selection

   • Choose an algorithm based on your problem type:
     • Classification (predicting categories)
     • Regression (predicting continuous values)
     • Clustering (grouping similar items)

4. Training Process

   • The model learns patterns from the training data
   • It works by:
     • Making predictions on training data
     • Calculating error using a loss function
     • Adjusting its parameters to minimize the error
     • Repeating this process (iterations/epochs)

5. Validation

   • Test the model on validation data
   • Tune hyperparameters (model settings)
   • Check for overfitting (when model performs well on training but poorly on new data)

6. Evaluation

   • Test final model performance on the test set
   • Use appropriate metrics (accuracy, precision, recall for classification; MSE, MAE for regression)