What you'll learn:
- Learn what machine learning is
- Learn the basic principles of machine learning
- Learn what artificial neurons are
- Learn the difference between machine learning and traditional programming
- Apply supervised learning techniques using Python libraries like scikit-learn
- Implement unsupervised learning algorithms such as clustering
- Build reinforcement learning agents and train them in simulated environments
- Understand and evaluate machine learning models using real-world datasets
- Visualize and interpret model results to gain actionable insights
- Compare the strengths and weaknesses of SL, UL, and RL in different scenarios
- Develop machine learning pipelines from data preprocessing to model deployment
- Use Python to experiment with hyperparameter tuning and model optimization
- Build an RNN in PyTorch to predict hourly temperature from real weather data and visualize real vs predicted values.
- Use Cursor and AI prompting to turn ML specifications into runnable code, from data preprocessing to model training and evaluation.
This hands-on machine learning course builds a clear, practical foundation in the field, moving from core ideas to real implementations. We open with what machine learning is, how it works, and the main application areas, then introduce supervised learning through the most accessible gateway: linear regression. You’ll see the contrast between traditional programming and the machine-learning approach, and learn how gradient descent works in practice—comparing full-batch, mini-batch, and stochastic updates. From there we step into the artificial neuron and use it to implement linear regression with a single neuron, cementing intuition before scaling up. A short quiz consolidates the essentials on supervised, unsupervised, and reinforcement learning.
We then dive deeper into regression: the algebraic method to ground your math intuition, one-dimensional regression, and the ML training pipeline for linear models, culminating in a truly simple neural network and extensions to multiple dependent variables. Next, in classification, you’ll assemble a deep neural network, define the target classes, and practice training and testing a model—connecting theory to usable evaluation workflows.
When moving to sequential data, you’ll learn how to predict hourly temperature with an RNN in PyTorch. In a dedicated lesson built with Cursor and AI prompting, you start from a detailed specification and use Cursor to generate a complete, runnable script: you’ll fetch real historical weather from meteo API, preprocess time series with MinMaxScaler and sliding-window sequences, implement a custom PyTorch Dataset and DataLoader, and train a simple RNN to forecast the next hour from the last 24, then evaluate and visualize real vs predicted temperature. The lesson shows how to go from a clear prompt to a working time series model with Cursor and AI-assisted coding.
Shifting to unsupervised learning, you’ll first understand and then code both K-Means and DBSCAN, before applying dimensionality reduction techniques and putting them into practice to make high-dimensional data tractable and visualizable. Finally, you’ll explore reinforcement learning: from the core principles to classic control problems like CartPole and LunarLander, and even design a custom environment, learning to reason in terms of states, actions, rewards, and policies. The course wraps up with a concise conclusion to help you review key takeaways and plan your next steps. Throughout, the emphasis is on conceptual clarity, clean code, and repeatable workflows—so you don’t just learn ML, you learn to build with it.
