INSTANCE BASED MACHINE LEARNING
What is Machine Learning?
Machine learning is an application of artificial intelligence that
generates programs for computer to access data and analyze it.
Based on this analysis, the system automatically learns and
improvises itself. Machine Learning is the study of certain set of rules
using statistical and logical approach to solve a particular problem.
These set of rules are called Machine Learning Algorithms and are
used in machine learning for continuous development and
improvement of specific task. Main aim of machine learning is to
make machines learn automatically and work without human
interference and assistance.
Machine Learning is an approach to make life easier for human
beings by reducing their efforts.
Face Recognition, virtual personal assistants like amazon’s alexa and
Google home, filtering spam emails and malwares, online customer
support, search engine result refining, social media services are a few
applications of machine learning.
Types of Machine Learning Algorithms:
Based on the purpose they serve machine learning algorithms can
be classified as:
Supervised Learning
Unsupervised Learning
Reinforcement Learning
Instance Based Machine Learning Algorithm:
Also known as Memory based learning, Instance based learning is a
supervised classification learning algorithm that performs operation
after comparing the current instances with the previously trained
instances, which have been stored in memory. Its name is derived
from the fact that it creates assumption from the training data
instances.
Time complexity of Instance based learning algorithm depends
upon the size of training data. Time complexity of this algorithm in
worst case is O (n), where n is the number of training items to be
used to classify a single new instance.
To improve the efficiency of instance based learning approach,
preprocessing phase is required. Preprocessing phase is a data
structure that enables efficient usage of run time modeling of test
instance.
Advantage of using Instance based learning over others is that it has
the ability to adapt to previously unseen data, which means that
one can store a new instance or drop the old instance.
Various instance based learning algorithms:
K Nearest Neighbor (KNN)
Self-Organizing Map (SOM)
Learning Vector Quantization (LVQ)
K-Nearest Neighbor:
K nearest neighbors of a classifier are used to create a test instance
model.
Example using KNN neighbor:
1.Splitting data into test and training datasets
2.Building and training model with training data
3. Evaluating model's predictions against the test dataset
Self-Organizing Map:
SOM is a type of artificial neural network (ANN) that is used for
producing map (i.e. low dimensional, different representation of
input data).
Unlike other artificial neural networks, SOM applies competitive
learning (i.e. learning in which nodes compete for the right to
respond to a subset of the input data).
Operations performed in SOM:
Select random input
Compute winner neuron
Update neurons
Repeat for all input data
Classify input data
BEFORE AFTER
SOM is used for dimensionality reduction (i.e. reduction of random
variables in account by acquiring a set of principal variables).
Learning Vector Quantization:
Learning Vector Quantization is a competitive network using
supervised learning. It is a process for classification of pattern in
such a way that each output represents a class.
