2. This chapter covers:
• The Concept of Machine Learning
• Supervised Machine Learning
• Unsupervised Machine Learning
• Applications of Machine Learning
3. Machine learning is one of the major application of Artificial
Intelligence that delivers the systems with the competence to
acquire mechanically and progress with practice and with certain
programming skills.
Machine learning can be considered as one of the techniques to analyze
the data which helps in developing the models systematically. It is a
subcategory of artificial intelligence established on the information that
enterprises can obtain from data, identify patterns and help in decision
making with very less involvement of humans.
For example if, traffic movements at a busy road is to be forecasted, a machine learning algorithm can
be used with past knowledge about traffic and, If the knowledge acquired is successful then the future
traffic movement can be predicted in a better way.
4. MACHINE LEARNING IS…
MACHINE LEARNING IS PROGRAMMING COMPUTERS TO OPTIMIZE A PERFORMANCE CRITERION USING EXAMPLE
DATA OR PAST EXPERIENCE.
-- ETHEM ALPAYDIN
THE GOAL OF MACHINE LEARNING IS TO DEVELOP METHODS THAT CAN AUTOMATICALLY DETECT PATTERNS IN DATA,
AND THEN TO USE THE UNCOVERED PATTERNS TO PREDICT FUTURE DATA OR OTHER OUTCOMES OF INTEREST.
-- KEVIN P. MURPHY
THE FIELD OF PATTERN RECOGNITION IS CONCERNED WITH THE AUTOMATIC DISCOVERY OF REGULARITIES IN DATA
THROUGH THE USE OF COMPUTER ALGORITHMS AND WITH THE USE OF THESE REGULARITIES TO TAKE ACTIONS.
-- CHRISTOPHER M. BISHOP
6. MACHINE LEARNING IS…
Machine learning is about predicting the future based on the
past.
-- Hal Daume III
Training
Data
model/
predictor
past
model/
predictor
future
Testing
Data
7. MACHINE LEARNING
• Data mining: machine learning applied to “databases”, i.e. collections of data
• Inference and/or estimation in statistics
• Pattern recognition in engineering
• Signal processing in electrical engineering
• Induction
• Optimization
8. Some machine learning methods
The Machine learning algorithms are divided into supervised or unsupervised
learning.
• In supervised machine learning algorithms, the historical knowledge acquired is used to
the current data so as to forecast the future happenings. It acquires a set of rules by
a training data for analysis to develop a model so as to forecast the dependent variable
which is also known as target variable.
The model provides the target value for any new values of the predictors after sufficient
training. Then the machine learning algorithm compares the estimated value with the
actual value of the target variable. The difference of actual and estimated value is
which has to corrected so as alter the
model.
9. Supervised learning vs. unsupervised learning
Supervised learning: discover patterns in the data that
relate data attributes with a target (class) attribute.
These patterns are then utilized to predict the values of the
target attribute in future data instances
Unsupervised learning: The data have no target
attribute.
We want to explore the data to find some intrinsic structures
in them.
9
10. Supervised learning needs supervision to train the model, which is
similar to as a student learns things in the presence of a teacher.
Supervised learning can be used for two types of problems:
Classification and Regression.
Unsupervised learning is another machine learning method in which
patterns inferred from the unlabeled input data. The goal of
unsupervised learning is to find the structure and patterns from the
input data. Unsupervised learning does not need any supervision.
Instead, it finds patterns from the data by its own.
Unsupervised learning can be used for two types of problems:
Clustering and Association.
11. In unsupervised machine learning, the data is not prepared and categorized.
This has benefit of extracting the information and provides patterns from the
hidden and unstructured data. Though the proper outcome is not generated
can come up with interpretations from the data to define unseen patterns from
unstructured data
The algorithm which lies between supervised and unsupervised learning is the
semi-supervised algorithm as it has the data which can be either a structured
or unstructured data as training data. This algorithm provides an enhanced
learning with efficient accuracy.
Generally, semi-supervised learning is preferred in the case of the data
obtained is structured and requires skills and to develop a model. Otherwise
unstructured data does not any require accompanying resources.
12. Some use cases for unsupervised learning — more
specifically, clustering — include: Customer segmentation, or
understanding different customer groups around which to
build marketing or other business strategies. Genetics, for
example clustering DNA patterns to analyze evolutionary
biology.
13.
14.
15. Supervised Machine Learning
Different Types of Supervised Learning
• Regression. In regression, a single output value is produced using training data. ...
• Classification. It involves grouping the data into classes. ...
• Naive Bayesian Model. ...
• Random Forest Model. ...
• Neural Networks. ...
• Support Vector Machines.
• Decision trees
27. Classification Applications
Face recognition
Character recognition
Spam detection
Medical diagnosis: From symptoms to illnesses
Biometrics: Recognition/authentication using physical and/or behavioral
characteristics: Face, iris, signature, etc
...
30. Regression Applications
Economics/Finance: predict the value of a stock
Epidemiology (the branch of medicine which deals with the incidence,
distribution, and possible control of diseases and other factors relating to
health.)
Car/plane navigation: angle of the steering wheel, acceleration, …
Temporal trends: weather over time
…
35. Unsupervised learning applications
learn clusters/groups without any label
customer segmentation (i.e. grouping)
image compression
bioinformatics: learn motifs
…
36. Reinforcement learning
Reinforcement learning is an approach to machine learning in which
the agents are trained to make a sequence of decisions. The agent,
also called as an AI agent gets trained in the following manner:
The agent interacts with the environment and make decisions or
choices. For training purpose, the agent is provided with the
contextual information about the environment and choices.
The agent is provided with the feedback or rewards based on how
well the action taken by the agent or the decision made by the
resulted in achieving the desired goal.
37. Reinforcement learning
Similar to toddlers learning how to walk who adjust actions based on
the outcomes they experience such as taking a smaller step if the
previous broad step made them fall, machines and software agents
use reinforcement learning algorithms to determine the ideal
behavior based upon feedback from the environment.
The biggest characteristic of this method is that there is no
supervisor, only a real number or reward signal. Two types
of reinforcement learning are 1) Positive 2) Negative.
Two widely used learning model are 1) Markov Decision Process 2) Q
learning
38. Self-driving cars: Reinforcement learning is used in self-driving cars for various
purposes such as the following. Amazon cloud service such as DeepRacer can
be used to test RL on physical tracks.
• Trajectory optimization
• Motion planning including lane changing, parking etc
• Dynamic pathing
• Controller optimisation
• Scenario-based learning policies for highways
39. Reinforcement learning
left, right, straight, left, left, left, straight
left, straight, straight, left, right, straight, straight
GOOD
BAD
left, right, straight, left, left, left, straight
left, straight, straight, left, right, straight, straight
18.5
-3
Given a sequence of examples/states and a reward after
completing that sequence, learn to predict the action to
take in for an individual example/state
40. Reinforcement learning example
… WIN!
… LOSE!
Backgammon
Given sequences of moves and whether or not the
player won at the end, learn to make good moves
41. Machine Learning Examples
One of the recent applications of machine leaning is Siri, Alexa, Google.
Based on the suggested name, it helps in obtaining material, when asked vocally. It can be
activated by asking certain questions and getting answers. In giving the answers Siri collects
related information and send communication to the other devices.
It can be given some instructions and obtain suggestions.
Machine learning is a substantial portion of the Alexa as it gathers and develops the evidences
based on the historical usage of it. Further, this information is used to excerpt outcomes that
are tailored to the favourites. They share the information with a variety of devices such as
Smart Speakers , Smart phones, Mobile Apps and so on.
42. Machine Learning Examples
Traffic Predictions:
We commonly use GPS, the direction-finding facility. This is achieved by saving, the existing
position and speed at a centralized server which manages the traffic. This data is later used to
provide a map giving the traffic situation at any point of time which supports in avoiding the
traffic by giving the analysis on the bottleneck places.
In these situations, Machine learning can be used to predict the crammed based on day-to-
day knowledge. Some other applications are such as booking a cab online , the app provides
the estimated price of the trip and also reduces the diversions with the support of machine
learning. It also helps in increasing the prices by forecasting the demand of the rides.
43. Machine Learning Examples
Machine learning is plays a significant role in the whole sequence of the services to be
provided. It helps in monitoring several cameras in Videos Surveillance as it is tough for only
one person to handle this.
Artificial Intelligence plays a crucial role which makes it feasible to sense the crime which may
happen. It is done by stalking uncommon behaviour of people . Hence it is possible for the
systems to report the uncommon activities so as to avoid any catastrophes.
This helps in improving the surveillance services and all of these are achieved with the help of
Machine learning.
Google and other search engines use machine learning algorithms to improve the search
results. The algorithms keep track on an individual responds to the results every time a search
is executed. The search engine understands the outcomes are in accordance with the request
if any result is opened for a long time.
44. Machine Learning Examples
Machine learning helps in enhancing the shopping experience on the products purchased
online by suggestions for future shopping to the customers based on the
products purchased.
The product recommendations are made based on the customer behavior with the
website/app, through the previous buying patterns on what are the items preferred liked or
purchased and the choice of the brand .
Machine learning is also providing capabilities in terms of tracking the online fraudulent
money transactions by using certain tools to identify authorized or unauthorized dealings
between the purchasers and venders.
45. Machine Learning Examples
Under supervised ML there are two most important subgroups
In regression the value of predicted variable is a continuous one which helps in
answering the required quantity and the numbers.
In classification, a regression predictor is not very useful. What we usually want is a predictor
that makes a guess somewhere between 0 and 1. It gives a function that
captures this behavior It’s called the sigmoid function.
Clustering divides the data into groups based on the similarity with respect to a certain
characteristic. Unsupervised machine learning is widely used in clustering.
Clustering tries to find several subcategories in a dataset. Since this is unsupervised learning,
there is no constraint on the number of labels and the choice of the number of clusters. This is
considered as pro or con since identification of the exact number of clusters has to be done
empirically.
46. Machine Learning Examples
In dimension reduction the purpose is to reduce the interrelated independent
variables into factors which are efficient in terms of converting the actual
information. If there are more variables or features which may be insignificant
Machine Leaning algorithms become less efficient. By reducing the dimensions
the major components are identified and used. The most common technique used
in this case is Principal Component Analysis (PCA).