Phase 2 · Math, when you need it · Probability & statistics
Maximum likelihood
Choose the parameters that make the data most probable.
Open this lesson Create a free account
Why it matters
How do you pick a model's parameters from data? The most common answer in all of statistics and machine learning is: choose the parameters that make the data you actually saw most probable. That's maximum likelihood. A conversion rate estimated as conversions ÷ visitors is a maximum-likelihood estimate, and so is a language model trained on the internet: its training objective, minimizing cross-entropy, is maximum likelihood in disguise.
What you'll be able to do
- Define the likelihood of data under a model's parameters
- Find the maximum-likelihood estimate by searching and by formula
- Work with log-likelihoods instead of likelihoods
- Explain why training a language model is maximum likelihood
Before this lesson
Next in this phase: Softmax