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

Evaluation
Essentially, in the context of ML "evaluation", is referred to as the method
of assessing various hypotheses or models to select one model over another.
An "evaluation function" is needed to distinguish between effective
classifiers from the vague ones. The evaluation function is also known as

---

the "objective," "utility," or "scoring" function. The machine-learning
algorithm has its internal evaluation function that is usually very different
from the researchers' external evaluation function used to optimize the
classifier. Usually, the evaluation function is described as the first phase of
the project before selecting the data representation tool. For example, the
self-driving car machine learning model has the feature to identify
pedestrians in its vicinity at near-zero, false-negative, and low false-positive
rate as an "evaluation function" and the pre-existing condition that needs to
be "represented" using applicable data features.
Optimization
The process of searching the hypothesis space of the represented machine
learning model to identify the highest-scoring classifier and achieve better
evaluation is called "optimization." For algorithms with more than one
optimum classifier, selecting the optimization method is very critical in
determining the generated classifier and achieving a more effective model
of learning. There are a variety of "off-the-shelf optimizers" on the market
to kick off new machine learning models before replacing them with
custom-designed optimizers.
Statistical Learning Framework
“Statistical learning” is a descriptive statistics-based learning framework
that can be categorized as supervised or unsupervised. “Supervised
statistical learning” includes constructing a statistical model to predict or
estimate output based on single or multiple inputs, on the other hand,
“unsupervised statistical learning” involves inputs but no supervisory
output, but helps in learning data relationships and structure. One way of

---

understanding statistical learning is to identify the connection between the
“predictor” (autonomous variables, attributes) and the “response”
(autonomous variable), in order to produce a specific model which is
capable of predicting the “response variable (Y)” on the basis of “predictor
factors (X)”.
“X = f(X) + ɛ where X = (X1,X2, . . .,Xp)”, where “f” is an “unknown
function” & “ ɛ ” is “random error (reducible & irreducible)”.
Here are some fundamental concepts of Statistical Learning: