Machine Learning ‘23/’24 Michiel de Folter
Machine Learning
Lecture 1: Introduction
Applications of Machine Learning:
- Regression
- Binary classification
- Multilabel classification
- Ranking
- Sequence to sequence prediction
- Many to one prediction
- One to many prediction
- Many to many prediction
Training, Validation & Test set: subsets of the data
- Training set: Used for inferring rules and learning by an update rule.
- Validation set: Used to choose the best learning parameters for the model to use.
- Test set: Used to generalize the model to real-world appliances.
Stratification:
A way of making sure that the sample is representative of the population by the standards of
your choosing. (Be careful about manually selecting the population for selection biases)
Regression evaluation metrics:
Mean Absolute Error (MAE):
N
1
MAE= ∗∑ ¿ y n−^
y n∨¿ ¿
N n=1
Mean Squared Error (MSE):
N
1
∗∑ ( y n−^
2
MSE= y n)
N n=1
R-squared:
2 MSE
R =1−
Variance of the True values
Classification evaluation metrics:
, Machine Learning ‘23/’24 Michiel de Folter
TRUE
Positive Negative
Positive TP FP Recall
PREDICTED
Negative FN TN Negative
prediction
power
Sensitivity Specificity Accuracy
Lecture 2: Decision trees
Decision tree: A repetition of ‘if-else’ statements called decision rules.
Machine Learning
Lecture 1: Introduction
Applications of Machine Learning:
- Regression
- Binary classification
- Multilabel classification
- Ranking
- Sequence to sequence prediction
- Many to one prediction
- One to many prediction
- Many to many prediction
Training, Validation & Test set: subsets of the data
- Training set: Used for inferring rules and learning by an update rule.
- Validation set: Used to choose the best learning parameters for the model to use.
- Test set: Used to generalize the model to real-world appliances.
Stratification:
A way of making sure that the sample is representative of the population by the standards of
your choosing. (Be careful about manually selecting the population for selection biases)
Regression evaluation metrics:
Mean Absolute Error (MAE):
N
1
MAE= ∗∑ ¿ y n−^
y n∨¿ ¿
N n=1
Mean Squared Error (MSE):
N
1
∗∑ ( y n−^
2
MSE= y n)
N n=1
R-squared:
2 MSE
R =1−
Variance of the True values
Classification evaluation metrics:
, Machine Learning ‘23/’24 Michiel de Folter
TRUE
Positive Negative
Positive TP FP Recall
PREDICTED
Negative FN TN Negative
prediction
power
Sensitivity Specificity Accuracy
Lecture 2: Decision trees
Decision tree: A repetition of ‘if-else’ statements called decision rules.
