Please find below a list of ML methods that you can choose from for your ML project.

If your desired methods are not in the lists below, please get in touch with us on Slack. We could consider added those into the lists if they are touched upon in the lectures or assignment. 

Course book: mlbook.cs.aalto.fi


Classification methods

Name
 Map Loss SK-Learn class ML book
K-Nearest Neighbors
Piece-wise constant around training data points.
0/1 loss KNeighborsClassifier 3.13
Logistic Regression
Linear maps Logistic loss LogisticRegression 3.6

Decision Tree


maps represented by a signal flow chart ("tree") that takes in features and maps them to a prediction by executing a series of “ if/else” decisions.

 Gini impurity/   Information gain  3.10
SVC
 linear maps applied to transformed features Hinge loss 3.7, 3.9
multi-layer perceptron
Linear maps + linear/nonlinear activation functions
Logistic loss
3.11


Regression methods

Name
 Map Loss SK-Learn class ML book
Linear Regression
 linear maps
 squared error
 3.1
Polynomial Regression
polynomial maps
squared loss
 3.2
Huber Regression
 linear maps Huber loss 3.3
Ridge Regression
 linear maps Mean squared error loss regularised by L2-norm
 7.1, 7.4, 7.8.1. 3.4
Lasso Regression
linear maps
 squared error plus L1-norm of parameters
 3.4, 7.1
Decision Tree
 maps represented by a signal flow chart ("tree") that takes in features and maps them to a prediction by executing a series of
“if/else” decisions.
squared error
 -
Multi-layer Perceptron
 map represented by a signal flow chart or "artificial neural network"
 squared error
3.11

Clustering

Name
 SK-Learn class ML book
k-means
KMeans 8.1
DBSCAN
 DBSCAN 8.3


Feature Learning

Name
 SK-Learn class ML book
Principal component analysis
PCA 9.2

Last modified: Thursday, 31 March 2022, 4:15 PM