What is machine learning?
- Machine learning maps input X to output Y as:
\[ Y = f(X) \]
- Machine learning provides two major things:
- Predicion
- Feature selection
- Categorized into:
- Parametric
- Non-parametric
- Unsupervised learning
- Supervised learning
- Hypothesis-driven approach
- Which features are characterisic for the type of cells etc. you want to predict?
Main steps of machine learning
- Clean the data: correct, normalize, standardize etc.
- Identify features in the data (deep learning skips this step, it builds its own features)
- Machine learning model is fitted on the training and evaluated on an independent subset
How does machine learning work?
Five steps:
- Split data set into train, validation and test subsets
- Randomly assign 70 % to training and 30 % to test (approx.)
- Fit model in the train subset
- Validate model on validation subset
- Repeat steps 1-3 a number of times
- Test the accuracy of the optimized model on test subset
What is a hyperparameter?
- Machine learning design parameters which are set before the learning process starts
- E.g. the number of covariates to adjust the main variable x of interest for
Random Forest
- Bases predictions on TRUE/FALSE trees
- Makes predictions based on the information given by iterating through the tree
What is Deep Learning?
- Artificial neural networks with multiple layers