Starting and learning
Machine learning for beginners
Machine learning finds patterns in examples so a program can predict something new, such as a house price, whether an email is spam or which species a flower is. Beginners usually start with tables of numbers, where regression predicts a quantity and classification predicts a category.
How machine learning is used: examples for beginners
Filtering spam email
A classifier reads features of a message, such as its words, links and sender history, and predicts spam or not spam. It learns from messages people have already marked, so it can adapt as spammers change tactics. This is the classic first classification problem.
Real exampleGmail spam filtering with TensorFlowGoogle reported in 2019 that adding TensorFlow models to Gmail's existing protections blocked around 100 million additional spam messages every day.
Recommending what to watch next
Recommendation models predict which items a person is likely to choose from what they and similar people chose before. Large platforms usually generate a shortlist of candidates first, then rank it with a second model.
Real exampleDeep Neural Networks for YouTube RecommendationsGoogle researchers describe YouTube's recommender as a deep candidate generation model followed by a separate deep ranking model, with lessons from running it at scale.
Translating text
Translation models learn from large collections of sentences paired with their translations. A neural network reads the whole sentence and then generates it in the other language, word by word.
Real exampleGoogle Neural Machine TranslationGoogle announced in 2016 that the Google Translate mobile and web apps were using its neural machine translation system for all Chinese to English translations.
Identifying species from sounds
Classification works on audio and images as well as tables. A neural network turns a recording into a spectrogram, an image of sound, and predicts which species is calling.
Real exampleBirdNETBirdNET is an open-source project that uses deep learning to recognise over 11,000 species globally from their sounds, with an app for citizen scientists.
Can a beginner learn machine learning?
Yes. Most people start with no machine learning background; what you need is basic Python, comfort with averages and percentages, and patience with messy data. Learn Python first, then data handling, then supervised learning on small tables, and leave neural networks until those feel routine.
- Python basicsVariables, lists, loops and functions, written in a notebook so you can run each line and see the result.Learn it in Python
- Working with tablesLoad a CSV with pandas, handle missing values and plot columns before any model is involved.Learn it in Data Tools
- Just enough mathsFunctions, the dot product and the idea of an error you can minimise. You need these to read what a model is doing, not to write proofs.Learn it in Maths
- Your first modelsSplit data into training and test sets, fit linear regression and a decision tree with scikit-learn, and compare both with a simple baseline.Learn it in Machine Learning
- Neural networks, laterOnce tabular models make sense, move to images and text with PyTorch.Learn it in Deep Learning
Moving from beginner to a machine learning role
If you already work in the field
If you work in another field, your knowledge of its problems is an advantage. Pick a prediction problem from your own work with data you can access and use it as your practice project; a realistic first move is an analyst or data role where modelling grows out of reporting.
If you come from data science or ML
If you already write code or analyse data, skip the Python basics and go straight to scikit-learn, train and test splits and evaluation metrics. Spend your effort on two or three complete projects with honest baselines that you can explain in detail.
Roles that do this workData analystJunior data scientistMachine learning engineer
Tools you will use
| Tool | What you use it for |
|---|---|
| Python | The language used for nearly all machine learning tutorials, libraries and course exercises. |
| pandas | Loads, cleans and summarises tables so you can see the columns before modelling. |
| scikit-learn | Fits regression and classification models, splits data and scores predictions with one consistent interface. |
| Jupyter | Runs code step by step in notebooks so you can inspect data and results as you go. |
| Matplotlib | Plots distributions, scatter plots and errors so you can see what the model is doing. |
Public data to practise on
- Wine Quality: Chemical measurements of red and white wines with a tasting score, ideal for a first regression or classification model.
- Auto MPG: Car attributes such as weight, horsepower and cylinders, used to predict fuel consumption with regression.
Your first project
Download the Wine Quality dataset, predict the quality score with linear regression, and compare it with always predicting the average score using mean absolute error on a held-out test set.
Common questions
Can I learn machine learning without a maths degree?
Yes. School algebra, averages and a feel for probability are enough to start. Linear algebra and calculus become useful when you study how models train, and you can learn them alongside practice.
Should a beginner start with deep learning or classical machine learning?
Start with classical models such as linear regression, logistic regression and decision trees on tables. They train in seconds, are easy to inspect and build the habits of splitting, baselines and metrics that deep learning also needs.
Do I need a powerful computer to learn machine learning?
No. Classical models on tables run on any laptop, and free notebook services such as Google Colab provide a GPU when you reach neural networks.
What is the best first machine learning project?
A small, clean table with one clear target, such as house prices or penguin species. Predict it, compare your model with a simple baseline and write down what the model gets wrong.
Courses for these skills
- Data Tools: NumPy, pandas, Matplotlib and scikit-learnInspect numeric columns, missing values and sample-level tables.23 published lessons · all planned lessons open
- Practical Machine Learning in PythonLearn regression, loss functions and generalisation on measured features.51 published lessons · more lessons in production
- Deep Learning with PyTorchStudy CNNs, representations and image-model evaluation.Syllabus available · lessons open as produced
- Python for AI and Machine LearningBuild the basic programming skills used to handle image records.33 published lessons · all planned lessons open
New to programming or maths? Start with Python, or follow the machine learning path in order. Published lessons are free to read; compare Free and Pro.