Machine learning

(T)ask

Classification

Classification with missing inputs

Regression

Transcription

Machine translation

Structure output

Anomaly detection

Synthesis and sampling

Imputation of missing values

Denoising

Density estimation or PMF estimation

(P)erformance

Accuracy

Error rate

Cross validation

Overfitting - When the gap between the traning-error and test-error is too large

Underfitting - Not low enough traning-error

Regularization - any modifaction we make to a learning algortihem that is intended to reduce its generalization error but not its traning error

MSE with regularization
λ is a hyperparameter, λ forces the weights to become smaller


Screenshot 2020-12-04 at 23.12.51

(MSE) Mean square error Screenshot 2020-12-04 at 23.10.59
Screenshot 2020-12-04 at 23.11.06