ai.harri.at¶
A compact wiki that leads from statistical inference through machine learning to modern artificial intelligence. The pages form a learning path: each part builds on the one before, but every page is written to be readable on its own.
One model, three perspectives
Logistic regression runs through the whole wiki. Statisticians use it to estimate how predictors affect the odds of a binary outcome; machine learning treats it as a baseline classifier; and a single artificial neuron with a sigmoid activation computes exactly the same function. Following this one model across the three parts is a good way to see how the fields connect.
The learning path¶
| Part | Central question | Start |
|---|---|---|
| 1. Statistical Inference | What can a sample tell us about a population, and how uncertain is that conclusion? | Statistical Inference |
| 2. Machine Learning | How do we learn a model from data that predicts well on new data? | Machine Learning |
| 3. AI | How do neural networks, transformers and language models work? | AI |
Key terms are collected in the glossary.
Prerequisites¶
The pages assume school-level mathematics: functions, sums, logarithms and a little probability. Some sections use vectors and matrices; they can be skimmed without losing the main ideas.
Mathematical notation¶
Formulas are typeset in the browser. Scalars are written in italics (\(x\)), vectors in bold lower case (\(\mathbf{x}\), \(\boldsymbol\beta\)) and matrices in bold upper case (\(\mathbf{X}\)). As an example, the density of the normal distribution with mean \(\mu\) and standard deviation \(\sigma\) is
Work in progress
This is an early version of the wiki. Pages will be extended and revised over time.