Machine learning – Basic concepts – Flashcards
Unlock all answers in this set
Unlock answersquestion
Classifier
answer
Composed of case base and learning algorithm to categorize real world objects
question
Case base consists of
answer
- O: set of variables - K: set of classes - E: set of cases
question
Case types
answer
- training: build internal representation of system - test: test performance on unseen data - start-up: require minimal prior knowledge - counterexamples: limit overgeneration
question
Supervised learning
answer
A desired output result for each input vector is required when the algorithm is trained
question
Unsupervised learning
answer
- The training of the learning algorithm is entirely data-driven, no need for target results for the input data vectors - cluster input data, find features inherent to the problem
question
Learning algorithm consists of
answer
- type: learning, genetic, neural network - supervised vs. unsupervised - knowledge representation scheme - classification vs. regression - presence of trainer
question
Classification trees
answer
- produce discrete output - split dataset based on homogeneity of data
question
Regression trees
answer
- prediction - needed when the response variable is numeric or continuous - At each split point, error is squared for "Sum of Squared Errors"
question
Inductive learning
answer
- x: input vector of attributes - f: function of value of target variable - want to learn mapping of x to f
question
Learning boolean functions
answer
- given examples of the function, can we learn it - boolean function: truth table + binary target function
