Machine Learning: 2 Books in 1: Machine Learning for Beginners, Machine Learning Mathematics. An Introduction Guide to Understand Data Science Through the Business Application

LASSO Regression. LASSO regression is another ‘shrinkage’ technique. A
very similar approach to ridge regression in that it encourages leaner,
simpler models for prediction. In lasso regression, the model is a little more
stringent about reducing the value of coefficients. LASSO stands for the
least absolute shrinkage and selection operator.
Data on our scatterplot is shrunk down to a more compact point, like the
average of the data. Just like ridge regression, we use this when the model
is suffering from multicollinearity.
ElasticNet Regression. ElasticNet regression works by combining the
techniques of LASSO and ridge regression. Its main goal is to attempt to
improve upon LASSO regression. It is a combination of both the methods
of rewarding lower coefficients in LASSO and Ridge regression. All three
of these models can be accessed in the glmnet package in R and Python.
Bayesian Regression. Bayesian regression models are helpful when we
have insufficient data or data with poor distribution. These types of
regressions are created using probability distributions rather than data
points, which means that the graph will appear as a bell curve representing
the variance with the most frequently occurring values in the middle of the
curve.
In Bayesian regression, the dependent variable Y is not a value but a
probability. Rather than trying to predict a value, we are trying to predict
the probability of an outcome. This is known as frequentist statistics, and
Bayes theorem makes up the foundation for this type of statistics.
Frequentist statistics hypothesize whether something will happen, and the
probability that it will happen.

---

When we talk about frequentist statistics, we also include conditional
probability. Conditional probability involved events whose outcomes are
dependent on one another. Every time you toss a coin, it is an independent
event meaning that the previous coin toss does not change the likelihood of
the next coin toss. Flipping a coin toss, therefore, is not a conditional
probability.
Events can also be dependent, meaning that the previous event can change
the probability of the next event. Say I had a bag of marbles, and I wanted
to know the probability drawing different colors out of the bag. If I have a
bag with 3 green marbles and 3 red marbles, and I draw a red marble, then
the probability of drawing a red marble goes down on my next draw. This
would be an example of conditional probability.