Unsupervised Machine Learning
It's used when we can't label the data, so the machine should work by it's own to find the patterns.
How it generally works
It's try to find patterns in the data, and then group the data based on the patterns.
- Gaussian initialization
- Expectation step
- Maximization step
- Repeat steps 2 and 3 until convergence
Expectation step
It's calculate the probability of each data point belonging to each cluster.
Maximization step
It's update the parameters of the model to maximize the likelihood of the data.
Convergence
It's when the algorithm stops updating the parameters.
When to use
- When we don't have labeled data
- When we want to find patterns in the data
- When we don't have data pre-existing about the result wanted