Glossary ML Flashcards
(33 cards)
K-nearest neighbour
Class predicition by majority vote
Overfitting
Modeling error due to too complex model
Weak learner
Classifier performing better than random guess
Latent variable
Variable inferred from other variables
Error backpropagation
Gradient following in a neural network/An approach to train artificial neural networks
Reward
Feedback from the environment
Variance of a classifier
Divergence of estimated prediction function
Fitness
Value used for the selection process
Support vector
Samples on the margin of the decision surface/Data point affecting the decision boundary
Negative sample
Training data which is not part of the concept being learned
k-means
Clustering method based on centroids
Normal distribution
Continous PDF defined by mean vector and covariance matrix
Posterior probability
Probability after observation/Conditional probability taking into account the evidence
The Lasso
An approach to regression that results in variable selection
Principal Component Analysis
An unsupervised method for dimensionality reduction
Perceptron learning
Error driven method to compute weights in a single layer neural network/Method to find seperating hyperplanes
Categorical distribution
Distribution of discrete stochastic variables
Subspace
A space spanned by a set of linearly independent vectors
Dropout
A method of regularization used in deep neural networks/An approach to train artificial neural networks
EM-algorithm
An approach to iteratively fitting model parameters with latent variable
Variance
Measure of spread of a random variable
Random Forests
Ensemble of decision trees
RANSAC
Robust method to fit a model to data with outliers
Curse of dimensionality
Issues in data sparsity
