Machine Learning

is machine learning AI?

Machine learning is a subset of AI

what is hypermarameter

a parameter whose value must be set by the researcher before learning begins

what is labeled data

also called training data; used in supervised learning

categories of supervised learning

1)regression model (output is predicted using training data); 2)classification problem (output is sorted into distinct group, like spam emails).

unsupervised learning

1) does not use labeled data 2) a set up inputs (x) is used for analysis with no corresponding target (y) 3) the algorithm discovers underlying structure in data

dimension reduction

reducing the number of input, complexity of data. Kinda like if I eat and work I carpool and eat I have less to do

clustering

reducing data by categorizing them. Example, putting kids going to soccer and going to piano together and then drive them. * observation with a cluster are similar and different accross clusters.

deep learning

self teaching system which a computer learns from interacting with itself. can be supervised, unsupervised or reinforcement learning (train and error)

overfitting

the ML fits the training data too well, unable to generalize new data

descriobe in sample error

1) regarding training sample 2)bias arises from underfitted models

out of sample errors

1) prediction erros in validation and text sample 2) variance error from over fitted models

residual errors

results from randomness in the data

complexity reduction and cross validation solves

overfitting

a type of cross validation — K fold

1) split the data set into K number of sections/folds 2) first section is used to test the model, and the test are used to train model 3) this reduces the problem of holdout sample (data not used to train the sample). * usually K = 5 or 10 (sections)

Lasso regression

1) is an example of penalty regression (penilize for having too many variable) 2) contracts the regression coefficient to 0

support vector machine

an algorithm for classification, regression and outlier detection. It draws a line (3D) between data to separate them