See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/342880643

Machine Learning Trends and Its Applications

Presentation · July 2020

DOI: 10.13140/RG.2.2.21226.75209

CITATIONS READS

0 1,746

1 author:

Hamed Daei Kasmaei

Islamic Azad University Central Tehran Branch
65 PUBLICATIONS 548 CITATIONS

SEE PROFILE

All content following this page was uploaded by Hamed Daei Kasmaei on 12 July 2020.

The user has requested enhancement of the downloaded file.

 Machine Learning Trends and Its
Applications
Lecturer : Prof.Dr.Hamed Daei Kasmaei
1Department of Mathematics and Statistics, Faculty of Science,
Central Tehran Branch, Islamic Azad University, Tehran, Iran.
2 Bijective Physics Group, Slovenia.
Email: [email protected]
[email protected]
Sponsor :
BABA FARID COLLEGE
Department of Computer Science
Two Days International Webinar on :
Artificial Intelligence and Machine Learning
Simulation tools and Softwares for AI/ML/DP
Weka: Workbench For Machine Learning
https://www.cs.waikato.ac.nz/ml/weka/
https://opensourceforu.com/2017/01/an-introduction-to-weka/
https://neurons.ai/directory/algorithms/software-tools/
https://www.softwaretestinghelp.com/machine-learning-tools/
https://scikit-learn.org/stable/
https://machinelearningmastery.com/what-is-the-weka-machine-learning-
workbench/
https://thenewstack.io/six-of-the-best-open-source-data-mining-tools/
https://github.com/matlab-deep-learning
Also: MATLAB,R, Python , Tensor Flow and Pythorch for Python
Good Books to Study by available full Text
Probability, Statistics and Estimation: Propagation of Uncertainties
in Experimental Measurement
Good Books to Study by available full Text
MATLAB Deep Learning: With Machine Learning, Neural Networks
and Artificial Intelligence
Good Books to Study by available full Text
Machine Learning Mastery with R
Good Books to Study by available full Text
An Introduction to Statistical Learning with Applications in R
Gareth James, Daniela Witten, Trevor Hastie and Robert Tibshirani
Winner of the 2014 Eric Ziegel award from Technometrics.
 What are neural networks?
Neural networks were vaguely inspired by the inner workings of
the human brain. The nodes are sort of like neurons, and the
network is sort of like the brain itself. (For the researchers among
you who are cringing at this comparison: Stop pooh-poohing the
analogy. It’s a good analogy.) But Hinton published his
breakthrough paper at a time when neural nets had fallen out of
fashion. No one really knew how to train them, so they weren’t
producing good results. It took nearly 30 years for the technique
to make a comeback. And boy, did it make a comeback.
What are neural networks?
A neural network is a series of algorithms that endeavors to
recognize underlying relationships in a set of data through a
process that mimics the way the human brain operates. In this
sense, neural networks refer to systems of neurons, either
organic or artificial in nature. Neural networks can adapt to
changing input; so the network generates the best possible
result without needing to redesign the output criteria. The
concept of neural networks, which has its roots in artificial
intelligence, is swiftly gaining popularity in the development of
trading systems.
What are neural networks?
A single-layer neural network represents the most simple form of
neural network, in which there is only one layer of input nodes
that send weighted inputs to a subsequent layer of receiving
nodes, or in some cases, one receiving node. This single-layer
design was part of the foundation for systems which have now
become much more complex.
What are neural networks?
Single layer NNs
Procedures of a Single-layer Perceptron Network
A perceptron will either send a signal, or not, based on the weighted inputs.
What is Artificial Intelligence (Definition)?
Artificial intelligence (AI), the ability of a digital computer or
computer-controlled robot to perform tasks commonly associated
with intelligent beings. The term is frequently applied to the project
of developing systems endowed with the intellectual processes
characteristic of humans, such as the ability to reason, discover
meaning, generalize, or learn from past experience. Since the
development of the digital computer in the 1940s, it has been
demonstrated that computers can be programmed to carry out very
complex tasks—as, for example, discovering proofs for
mathematical theorems or playing chess—with great proficiency.
Artificial intelligence is the broader concept that consists of
everything from Good Old-Fashioned AI (GOFAI) all the way to
futuristic technologies such as deep learning.
WHY AI/ML/DP ON DATA?
Data for AI
 MAIN FORMULAS FOR MACHINE LEARNING
ML in one Look
What is Artificial Intelligence (Definition)?
•As the name suggests, artificial intelligence can be loosely
interpreted to mean incorporating human intelligence to
machines.
•Still, despite continuing advances in computer processing speed
and memory capacity, there are as yet no programs that can
match human flexibility over wider domains or in tasks requiring
much everyday knowledge. On the other hand, some programs
have attained the performance levels of human experts and
professionals in performing certain specific tasks, so that artificial
intelligence in this limited sense is found in applications
as diverse as medical diagnosis, computer search engines, and
voice or handwriting recognition.
 Note:
 Whenever a machine completes tasks based on a set of stipulated
rules that solve problems (algorithms), such an “intelligent” behavior
is what is called artificial intelligence.
 For example, such machines can move and manipulate objects,
recognize whether someone has raised the hands, or solve other
problems.
 AI-powered machines are usually classified into two groups —
general and narrow. The general artificial intelligence AI machines
can intelligently solve problems, like the ones mentioned above.
 The narrow intelligence AI machines can perform specific tasks very
well, sometimes better than humans — though they are limited in
scope.
 The technology used for classifying images on Pinterest is an
example of narrow AI.(As a social media analysis).
What is Machine Learning Definition ?
Machine-learning algorithms use statistics to find patterns in
massive amounts of data. And data, here, encompasses a lot of
things—numbers, words, images, clicks, what have you. If it can
be digitally stored, it can be fed into a machine-learning
algorithm.
Machine learning is the process that powers many of the services
we use today—recommendation systems like those on Netflix,
YouTube, and Spotify; search engines like Google and Baidu;
social-media feeds like Facebook and Twitter; voice assistants
like Siri and Alexa. The list goes on.
 What is Machine Learning Definition ?
 In all of these instances, each platform is collecting as much
data about you as possible.
what genres you like watching, what links you are clicking,
which statuses you are reacting to—and using machine
learning to make a highly educated guess about what you
might want next. Or, in the case of a voice assistant, about
which words match best with the funny sounds coming out of
your mouth. Main Process:
 Frankly, this process is quite basic: find the pattern, apply the
pattern. But it pretty much runs the world. That’s in big part
thanks to an invention in 1986, courtesy of Geoffrey Hinton,
today known as the father of deep learning.(Also Alan Turing
Efforts).
What is Machine Learning Definition ?
 As the name suggests, machine learning can be loosely
interpreted to mean empowering computer systems with the
ability to “learn”.
 The intention of ML is to enable machines to learn by
themselves using the provided data and make accurate
predictions.ML is a subset of artificial intelligence; in fact, it’s
simply a technique for realizing AI.
 It is a method of training algorithms such that they can learn
how to make decisions.
 Training in machine learning entails giving a lot of data to the
algorithm and allowing it to learn more about the processed
information.
What is deep learning?

Deep learning is machine learning algorithm that uses a

technique at which gives machines an enhanced ability to find—
and amplify—even the smallest patterns. This technique is called
a deep neural network—deep because it has many, many layers
of simple computational nodes that work together to munch
through data and deliver a final result in the form of the
prediction.
Difference Between ML and DP
ML VS DP
Differences Between Machine Leaning (ML),
Artificial Intelligence (AI) and Deep Learning
(DL)
AI vs Machine Learning vs Deep Learning Differences
Here is an image that attempts to visualize the distinction
between them:
Differences Between Machine Leaning (ML),
Artificial Intelligence (AI) and Deep Learning(DP)
 What this picture says to us ?
 AI means getting a computer to mimic human
behavior in some way.
 Machine learning is a subset of AI, and it consists of
the techniques that enable computers to figure
things out from the data and deliver AI
applications.
 Deep learning, meanwhile, is a subset of machine
learning that enables computers to solve more
complex problems.
What is feature extraction/selection?
Straight to the point:
 Extraction: Getting useful features from existing data.
 Selection: Choosing a subset of the original pool of features.
Why must we apply feature extraction/selection?
 Feature extraction is a quite complex concept concerning the
translation of raw data into the inputs that a particular Machine
Learning algorithm requires. The model is the motor, but it
needs fuel to work. Features must represent the information of
the data in a format that will best fit the needs of the algorithm
that is going to be used to solve the problem.
 While some inherent features can be obtained directly from
raw data, we usually need derived features from these inherent
features that are actually relevant to attack the underlying
problem. A poor model fed with meaningful features will surely
perform better than an amazing algorithm fed with low-quality
features – “garbage in, garbage out”.
Why must we apply feature extraction/selection?

 Feature extraction fills this requirement: it builds valuable

information from raw data – the features – by reformatting,
combining, transforming primary features into new ones… until
it yields a new set of data that can be consumed by the
Machine Learning models to achieve their goals.
 Feature selection, for its part, is a clearer task: given a set of
potential features, select some of them and discard the rest.
Feature selection is applied either to prevent redundancy
and/or irrelevancy existing in the features or just to get a limited
number of features to prevent from overfitting.
Why must we apply feature extraction/selection?
 Feature extraction is for creating a new, smaller set of
features that stills captures most of the useful information.
Again, feature selection keeps a subset of the original
features while feature extraction creates new ones.
 As with feature selection, some algorithms already have built-
in feature extraction. The best example is Deep Learning,
which extracts increasingly useful representations of the raw
input data through each hidden neural layer.
Note:
 Note that if features are equally relevant, we could perform PCA
technique to reduce the dimensionality and eliminate
redundancy if that was the case. Here we would be doing
feature extraction, as we were transforming the primary features
and not just selecting a subset of them.
 PCA=(Principal Component Analysis)
 Principal component analysis (PCA) is an unsupervised
algorithm that creates linear combinations of the original
features. The new features are orthogonal, which means that
they are uncorrelated. Furthermore, they are ranked in order of
their "explained variance." The first principal component (PC1)
explains the most variance in your dataset, PC2 explains the
second-most variance, and so on.
For More Info, Please visit:
https://quantdare.com/what-is-the-difference-between-feature-extraction-
and-feature-selection/
Note: What are labels on data?
Labels can be obtained by asking humans to make judgments
about a given piece of unlabeled data (e.g., "Does this photo
contain a horse or a cow?"), and are significantly more
expensive to obtain than the raw unlabeled data.
After obtaining a labeled dataset, machine learning models
can be applied to the data so that new unlabeled data can be
presented to the model and a likely label can be guessed or
predicted for that piece of unlabeled data.
Regression and Classification problems
 Classification problems ask the algorithm to predict a discrete
value, identifying the input data as the member of a particular
class, or group. In a training dataset of animal images, that
would mean each photo was pre-labeled as cat, koala or
turtle. The algorithm is then evaluated by how accurately it can
correctly classify new images of other koalas and turtles.
 On the other hand, regression problems look at continuous
data. One use case, linear regression, should sound familiar
from algebra class: given a particular x value, what’s the
expected value of the y variable?
 A more realistic machine learning example is one involving
lots of variables, like an algorithm that predicts the price of an
apartment in San Francisco based on square footage, location
and proximity to public transport.
10 AI and Machine Learning Trends To
Impact Business in 2020
 AI Is Helping Combat COVID-19
 ML Framework Competition
 AI Analysis for Business Forecasts
 Reinforcement Learning
 AI-driven Biometric Security Solutions
 Automated Machine Learning
 Explainable AI
 Conversational AI
 Generative Adversarial Networks
 Convergence of IoT and AI
Trend 1. AI Is Helping Combat COVID-19
A World Health Organization report from February 2020 revealed
AI and big data are playing an important role in helping
healthcare professionals respond to the coronavirus (COVID-19)
outbreak in China.
So, how is AI and machine learning helping combat COVID-19?
There are many applications, including:
Thermal cameras and similar technologies are being used to read
temperatures before individuals enter busy places like public
transport systems, government buildings, and other important
areas. In Singapore, one hospital is leveraging KroniKare’s
technology to provide on-the-go temperature checks using
smartphones and thermal sensors.
Trend 1. AI Is Helping Combat COVID-19
Chinese tech company Baidu created an AI system that uses
infrared technology to predict passengers temperatures at
Beijing’s Qinghe Railway Station.
Robots are being deployed to implement “contactless
delivery” for isolated individuals, helping medical staff ensure
that key areas stay disinfected and safe for use.
Many diagnosis algorithms have been developed around the
world along with creating vaccines in many institutes and
research centers and it will be expected that AI can be one of
factors that specifies and detects many behaviors of COVID
19.
Trend 1. AI Is Helping Combat COVID-19
E-commerce giant Alibaba created the StructBERT NLP model to
help combat COVID-19. This platform provides healthcare data
analysis using the company’s existing platforms and search
engine capabilities, which proved instrumental in expediting the
country’s ability to disseminate medical records.
Solutions like these provide a proactive approach to threat
detection, which can limit the spread of infectious diseases. And
when it comes to something as contagious as COVID-19, a pro-
active approach isn’t just important—it’s essential.
Trend 2. ML Framework Competition
In 2019, one of the key trends in the ML was PyTorch vs. TensorFlow
competition. During 2019, TensorFlow 2 arrived with Keras integrated
and eager execution default mode. PyTorch eventually overtook
TensorFlow as the framework of choice for AI research.
Why is PyTorch better for research? PyTorch integrates easily with the
rest of Python. And it is simple and easy to use, making it accessible
without requiring too much effort to set it up. In contrast, TensorFlow
crippled itself by repeatedly switching APIs, making it more difficult to
use.
When it comes to performance, PyTorch has comparable speed to
TensorFlow, which makes it technologically superior. Still, TensorFlow is
compatible with more business solutions, though, so most businesses
have not made the switch yet. While PyTorch is now the common
framework used for research, businesses are still using TensorFlow well
into 2020.
Trend 3. AI Analysis for Business Forecasts
ML-based time series analysis is a hot AI trend in 2020. This
technique collectively analyzes a series of data over time. When
used correctly, it aggregates data and analyzes it in such a way
that allows managers to easily make decisions based on their
data.
Using an ML network to process the complex calculations
required to apply statistical models to your business’s structured
data is a major improvement over traditional methods. This ML-
boosted analysis offers high-accuracy forecasts that are 90-95%
accurate. When the AI network you’re using is properly trained, it
can capture features of your business, such as seasonality and
cross-correlation in demand forecasting for retail.
Trend 3. AI Analysis for Business Forecasts
In 2020 we’ll see a growing trend for applying recurrent neural
networks for time series analysis and forecasting. Recurrent
neural networks, which are an application of deep learning, are
one reason we believe that deep learning will end up replacing
traditional machine learning. For example, deep learning can
forecast data, such as future exchange rates for currency with a
surprisingly high degree of accuracy.
The research into time series classification has made substantial
progress in recent years. The problem being solved is complex,
offering both high dimensionality and large numbers. So far, no
industry applications have been achieved. However, this is set to
change as the research into this field has produced many
promising results.
Trend 3. AI Analysis for Business Forecasts
Another type of artificial intelligence that has been recently
developed is the convolutional neural network (CNN). This type
of ML network discovers and extracts the internal structure that is
required to generate input data for time series analysis.
Along with forecasting the future, there’s another technology that
could be widely applied: anomaly detection based on
autoencoders that run artificial neural networks using
unsupervised learning algorithms. These systems are capable of
capturing common patterns while ignoring “noise.” Encoded
feature vectors allow businesses to separate anomalies, such as
financial, political, and even social data.
Trend 4. Reinforcement Learning
Reinforcement learning (RL) is leading to something big in 2020.
RL is a specialized application of deep learning that uses its own
experiences to improve itself, and it’s effective to the point that it
may be the future of AI.
When it comes to reinforcement learning AI, the algorithm learns
by doing. Initially, actions are tried at random, but eventually,
this becomes a logical process as it attempts to attain specific
goals. The operator rewards or punishes these actions, and the
results are fed back into the network to “teach” the AI.
No predefined suggestions are given to the reinforcement
learning agent. Instead, the AI starts out by acting completely
randomly, and eventually learns how to maximize its reward
through repetition. Reinforcement learning allows the algorithm
to develop sophisticated strategies.
Trend 4. Reinforcement Learning
 Reinforcement learning is the best way to simulate human
creativity in a machine by running many possible scenarios.
The model can even be adapted to complete complex
behavioral tasks. It’s an ideal solution for solving all kinds of
optimization problems.
 Self-improving chatbots are one example of reinforcement
learning’s effect. A goal-oriented chatbot is one that is
designed to help a user solve a specific problem, such as
making an appointment or booking a ticket to an event. A
chatbot can be trained using reinforcement learning through
trial and error to become a fully functional automated assistant
to customers.
Trend 5. AI-driven Biometric Security Solutions
 Significant advancements have been made in biometric
verification. Bio-ID is no longer something you’d expect to see in
sci-fi films. This emerging ML trend is one to keep your eye on.
 ML’s efficient approach to gathering, processing, and analyzing
large data sets can improve the performance of your biometric
systems. Running an efficient biometrics system is all about
performing matching tasks quickly and accurately, and this is a
task that ML networks excel at.
 The reliability of AI based biometric security is also increasing.
Here’s an example: a deep learning-based face anti-spoofing
system allows you to secure any face recognition solution from
any attempt to imitate a real face.(ANOTHER EXAMPLE: Carved
ID ON BORN CHILDREN IN SWEDEN).
Trend 5. AI-driven Biometric Security
Solutions
Another example of biometrics ML applications is Amazon’s
Alexa, which is now able to tell who is speaking by comparing
the speaker to a predetermined voice profile. No extra
hardware is necessary to help a properly trained neural network
to accurately identify the speaker.
In 2020, we predict that various biometrics will be combined
with ML to create a comprehensive security solution. Multimodal
biometric recognition is within our reach, thanks to
advancements in AI technology.
Trend 6. Automated Machine Learning
 AutoML is adapted to execute tedious modeling tasks that
once required weeks or months of work by professional data
scientists.
 AutoML runs systematic processes on the raw input data to
choose the model that makes the most sense. AutoML’s job is
to find a pattern in the input data and decide what model is
best applied to it. Previously these activities were processed
by hand.( Not Manual)
 AutoML applies several different machine learning
techniques. Google’s AutoML (a combination of recurrent
neural network (RNN) and reinforcement learning) is one
example. After extensive repetition, a high degree of
accuracy can be achieved automatically.
Trend 6. Automated Machine Learning
 Major cloud computing services offer a type of AutoML.
Google AutoML and Azure Automated Machine Learning are
two popular examples. Other options include the open-source
AutoKeras, tpot, and AutoGluon MLaaS platforms. The best
choice for your business will depend on your business’s goals
and budget.
 So, is AutoML effective? The answer in practice is often yes. For
example, Lenovo was able to use DataRobot by AWS to
reduce model creation time for their demand forecasts from 3-
4 weeks to 3 days—representing an impressive sevenfold
improvement. Model production time was lowered by an even
larger factor, all the way from two days to five minutes! The
prediction accuracy of these models has also increased.
Trend 7. Explainable AI
 The European Union tasked ML designers, known as the Right to
Explanation, to make artificial intelligence more transparent to consumers
and users. Explainable AI is a type of AI technology that has
been designed to fit these criteria.
 Unlike regular black-box machine learning techniques, where it’s often
impossible to explain how the AI came to a certain conclusion,
explainable AI is designed to simplify and visualize how ML
networks make decisions.
 What does “black box” mean? In traditional AI models, the network is
designed to produce either a numerical or binary output. For example, an
ML model designed to decide whether to offer credit in specific situations
will output either “yes” or “no,” with no additional explanation. The output
with explainable AI will include the reasoning behind any decision made
by the network, which using our example, allows the network to provide a
reason for approving or denying the credit request.
Trend 7. Explainable AI
 Businesses are starting to rely on various trending machine
learning algorithms to make decisions. According to Gartner,
around 30% of large enterprise contracts are likely to require
these solutions by 2025. Explainable AI is necessary if
companies require proper accountability during these
processes.
 One example of this future trend in AI is Local Interpretable
Model-Agnostic Explanation (LIME). This Python library
explains the predictions of any classifier by learning a special
human-readable model around the predictions. With LIME
and other techniques, even non-experts in the field are able
to find and improve inaccurate models. This is still a very new
field with plenty of room for improvement.
Trend 8. Conversational AI
 Throughout 2019 and 2020, artificial intelligence has
developed to a point where it can now compete with the
human brain when it comes to everyday tasks, such as writing.
Researchers at OpenAI claim that their AI-based text
generator is able to generate realistic stories, poems, and
articles. Their GPT-2 network was trained using a large writing
data set and can adapt to different writing styles on demand.
 Bidirectional Encoder Representations from Transformers (BERT)
is another significant outcome in the AI field. This is another
text AI that is designed to pre-train models using given text.
The major advancement is how BERT processes text.
 Trend 8. Conversational AI
Unlike previous approaches, which read the text either from left to
right or right to left, but never both, BERT brings a language model that
allows for bidirectional training. BERT has a deeper understanding of
language than any network that came before it and uses several
types of preceding architecture to generate accurate predictions for
text.(Like Microsoft Word Text Prediction)
The better the computer understands text that is fed into it, the higher-
quality the machine’s responses will be. BERT is a step closer to an AI
that is able to accurately understand and answer questions that are
fed into it, just like a human could.
XLNet is an autoregressive pre-training model that’s able to predict
words from a set of text using context clues. Despite being only a
simple feed-forward algorithm, it has managed to outperform BERT in
many NLP tasks.(Used as NLP Algorithms to process languages)
Trend 8. Conversational AI
 One clear application is voice-enabled AI. Voice activation
and voice commands all function on the basis of the
computer understanding the voice-to-text transcript of the
spoken command. The better the computer can understand
the text, the more accurately it can perform spoken
commands as well.
 With over 110 million virtual assistant users in the USA alone,
there is a massive market for improving voice recognition.
Today, voice-enabled devices, such as the Amazon Echo
and Google Home, are common in homes. Any
improvement to the voice assistant technology will lead to
an increase in business in this sector, and ML is the quickest
path to achieving these improvements.
Trend 9. Generative Adversarial Networks
Generative Adversarial Networks are a way to generate new
data using existing data in such a way that the new product
resembles the original. This may not seem too impressive at
first—after all—copying is easy, right? Well, not quite.
By generating similar but non-identical data, GANs are able to
produce amazing data, such as synthetic photos of a human
face that are indistinguishable from a real human.
Since being invented by Ian Goodfellow in 2014, GANs have
achieved significant progress in the field of synthetic face
generation.
 Trend 9. Generative Adversarial Networks
There is an impressive example of GAN technology at work. A fake
face generator was developed by Nvidia. It’s known as This Person
Does Not Exist, and has gained some traction online. Other examples
are facial processing apps that can produce aged or gender-swapped
versions of an existing photo.
So, how do GANs work? A GAN is an ML network that is trained using
two neural network models: a generator model and a discriminator
model. One of these models, the generator, is responsible for creating
new data samples. The discriminator’s job is to decide whether the
generated data is distinguishable from real data samples. During
training, the two models are competing against each other, with the
generator trying to fool the discriminator.
But it’s not all perfect. Advancements in GANs have caused concern in
the industry through their ability to synthesize totally fabricated, but
realistic images. The DeepFakes scandal is an example of how GANs
can be misused.
Trend 9. Generative Adversarial Networks

A properly used GAN is able to generate new images from only

a description. Soon enough, these networks will be used for
applications such as police sketches. Aside from generating
new images, the discriminator of a GAN is a good way to
detect anomalies and holds plenty of applications in quality
control and other inspection-based work.
 Trend 10. Convergence of IoT and AI
 Industrial IoT processes are generally not as efficient as they
could be. This leaves plenty of room for AI algorithms to help
increase efficiency and reduce downtime for various
businesses through methods such as predictive maintenance
or defect detection. Overall, the addition of AI to a
manufacturing process can only increase its efficiency.
 The current IoT trends reveal that businesses are accepting the
potential of ML. Rolls Royce partnered with Azure IoT Solutions
to use the cloud and IoT devices to their advantage. The
power of predictive maintenance shouldn’t be
underestimated, and Rolls Royce is taking advantage of their
IoT devices to check the health of their aircraft engines to
keep their uptime at a maximum.
 Trend 10. Convergence of IoT and AI

 Another company that has jumped onto the IoT-AI bandwagon

is Hershey. In Hershey’s production facilities, even a 1%
variance in weight can cost a lot. Using Microsoft Azure
machine learning network, Hershey was able to significantly
reduce the variability of their product weight, resulting in major
savings.
 One important machine learning option to improve the IoT
software development process is knowledge distillation.
Trend 10. Convergence of IoT and AI
Mechanism:
Trend 10. Convergence of IoT and AI
 In this type of learning, ML network learns how to produce desired results
through techniques such as reinforcement learning. Then, a small ML
network is trained to produce identical results to the large ML network. The
point of knowledge distillation is model compression. This smaller ML network
is easier to run on less powerful devices, such as IoT sensors and other
mobile devices. With knowledge distillation, it’s possible to decrease an ML
model’s weight on a given device by up to 2000%, saving both on energy
and hardware budget.
 An example of knowledge distillation is a video surveillance system that
needs to detect the genders of people on camera in real-time. Detecting a
person’s gender takes a large neural network, which is best run on the
cloud. However, for real-time detection, you can’t always rely on the cloud.
By distilling the larger network’s knowledge into the smaller one, the same
gender detection task can be performed by a network small enough to fit
onto a small mobile or edge device, resulting in large savings.
Trend 10. Convergence of IoT and AI
The Future of AI: It Is Only Getting Started
Advancements in hardware, computing power, and other
technical specifications will continue to fuel the rise of AI
technologies.
What is supervised learning?

Machine (and deep) learning comes in three flavors:

supervised, unsupervised, and reinforcement. In supervised
learning, the most prevalent, the data is labeled to tell the
machine exactly what patterns it should look for. Think of it as
something like a sniffer dog that will hunt down targets once it
knows the scent it’s after. That’s what you’re doing when you
press play on a Netflix show—you’re telling the algorithm to find
similar shows.
 With supervised machine learning, the
algorithm learns from labeled data.
Labelling data needs more human action and its not beneficial rather than
Automation.
What is supervised learning?
 We think of supervised learning with the concept of
function approximation, where basically we train an
algorithm and in the end of the process
we pick the function that best describes the input
data, the one that for a given X makes the best
estimation of y (X -> y).
 Most of the time, we are not able to figure out the true
function that always make the correct predictions and
other reason is that the algorithm rely upon an
assumption made by humans about how the computer
should learn and this assumptions introduce a bias.
What is supervised learning?
 Here the human experts acts as the teacher where we
feed the computer with training data containing the
input/predictors and we show it the correct answers
(output) and from the data the computer should be able to
learn the patterns.
 Supervised learning algorithms try to model relationships
and dependencies between the target prediction output
and the input features such that we can predict the output
values for new data based on those relationships which it
learned from the previous data sets.
 Applications of Supervised Learning
 Predictive Model.
 we have labeled data.
 The main types of supervised learning problems
include regression and classification problems.
 Nearest Neighbor (In Analyzing Scatted Data in Statistics)-KNN
 Naive Bayes (Classification Algorithms based on bayes
Theorem in Probability Courses)
https://www.analyticsvidhya.com/blog/2017/09/naive-bayes-explained/
 Decision Trees (In Graph Algorithms)
 Linear Regression (In Statistics)
 Support Vector Machines (SVM):To analyze data for classification and
Regression in order to train Data
https://monkeylearn.com/blog/introduction-to-support-vector-machines-svm/
 Neural Networks
Applications of Supervised Learning
Labeled data: Data consisting of a set of training
examples, where each example is a pair consisting of an input
and a desired output value (also called the supervisory signal,
labels, etc)
Classification: The goal is to predict discrete values, e.g. {1,0},
{True, False}, {spam, not spam}.
Regression: The goal is to predict continuous values, e.g. home
prices.
What is unsupervised learning?
•In unsupervised learning, the data has no labels. The machine
just looks for whatever patterns it can find. This is like letting a
dog smell tons of different objects and sorting them into groups
with similar smells. Unsupervised techniques aren’t as popular
because they have less obvious applications. Interestingly, they
have gained traction in cybersecurity.
•The computer is trained with unlabeled data.
•Here there’s no teacher at all, actually the computer might be
able to teach you new things after it learns patterns in data,
these algorithms a particularly useful in cases where the human
expert doesn’t know what to look for in the data.
What is unsupervised learning?
They are the family of machine learning algorithms which are
mainly used in pattern detection and descriptive modeling.
However, there are no output categories or labels here based
on which the algorithm can try to model relationships. These
algorithms try to use techniques on the input data to mine for
rules, detect patterns, and summarize and group the data
points which help in deriving meaningful insights and describe
the data better to the users.
Unsupervised learning models automatically
extract features and find patterns in the data.
Unsupervised Learning based on different patterns and datasets
 What is unsupervised learning?
Applications:
 Descriptive Model
 The main types of unsupervised learning algorithms
include Clustering algorithms and Association rule learning
algorithms.
 k-means clustering, Association Rules
What is reinforcement learning?
A reinforcement algorithm learns by trial and error to achieve
a clear objective. It tries out lots of different things and is
rewarded or penalized depending on whether its behaviors
help or hinder it from reaching its objective. This is like giving
and withholding treats when teaching a dog a new trick.
Reinforcement learning is the basis of Google’s AlphaGo, the
program that famously beat the best human players in the
complex game of Go.
 What is reinforcement learning?
Method aims at using observations gathered from the interaction
with the environment to take actions that would maximize the
reward or minimize the risk. Reinforcement learning algorithm
(called the agent) continuously learns from the environment in
an iterative fashion. In the process, the agent learns from its
experiences of the environment until it explores the full range of
possible states.
Reinforcement Learning is a type of Machine Learning, and
thereby also a branch of Artificial Intelligence. It allows machines
and software agents to automatically determine the ideal
behavior within a specific context, in order to maximize its
performance. Simple reward feedback is required for the agent
to learn its behavior; this is known as the reinforcement signal.
What is reinforcement learning?
What is reinforcement learning?
There are many different algorithms that tackle this issue. As a
matter of fact, Reinforcement Learning is defined by a specific
type of problem, and all its solutions are classed as
Reinforcement Learning algorithms. In the problem, an agent is
supposed decide the best action to select based on his current
state. When this step is repeated, the problem is known as
a Markov Decision Process.
In order to produce intelligent programs (also called agents),
reinforcement learning goes through the following steps:
What is reinforcement learning?
Steps:
Input state is observed by the agent.
Decision making function is used to make the agent perform an action.
After the action is performed, the agent receives reward or reinforcement
from the environment.
The state-action pair information about the reward is stored.
What is reinforcement learning?
List of Common Algorithms
Q-Learning
Temporal Difference (TD)
Deep Adversarial Networks

Use cases:
Some applications of the reinforcement learning algorithms are
computer played board games (Chess, Go), robotic hands,
and self-driving cars.
What are Convolutional Neural Networks?
 A Convolutional Neural Network (ConvNet/CNN) is a Deep
Learning algorithm which can take in an input image, assign
importance (learnable weights and biases) to various
aspects/objects in the image and be able to differentiate
one from the other. The pre-processing required in a ConvNet
is much lower as compared to other classification algorithms.
While in primitive methods filters are hand-engineered, with
enough training, ConvNets have the ability to learn these
filters/characteristics.
 The architecture of a ConvNet is analogous to that of the
connectivity pattern of Neurons in the Human Brain and was
inspired by the organization of the Visual Cortex. Individual
neurons respond to stimuli only in a restricted region of the
visual field known as the Receptive Field. A collection of such
fields overlap to cover the entire visual area.
A CNN sequence to classify handwritten digits
Applications of Convolutional Neural Networks
 Decoding Facial Recognition
 Analyzing Documents
 Historic and Environmental Collections
 Understanding Climate
 Grey Areas(Better Image resolution about what human being
sees)
 Advertising
 brain cancer detection and many issues in Healthcare
Recurrent neural networks
 A recurrent neural network (RNN) is a type of artificial neural
network commonly used in speech recognition and natural
language processing (NLP). RNNs are designed to recognize a
data's sequential characteristics and use patterns to predict
the next likely scenario.
 RNN converts the independent activations into dependent
activations by providing the same weights and biases to all the
layers, thus reducing the complexity of increasing parameters
and memorizing each previous outputs by giving each output
as input to the next hidden layer.
 Hence these three layers can be joined together such that the
weights and bias of all the hidden layers is the same, into a
single recurrent layer.
Recurrent neural networks

 RNN have a “memory” which remembers all information about what

has been calculated. It uses the same parameters for each input as it
performs the same task on all the inputs or hidden layers to produce
the output. This reduces the complexity of parameters, unlike other
neural networks.
 RNNs are used in deep learning and in the development of models
that simulate the activity of neurons in the human brain. They are
especially powerful in use cases in which context is critical to
predicting an outcome and are distinct from other types of artificial
neural networks because they use feedback loops to process a
sequence of data that informs the final output, which can also be a
sequence of data . These feedback loops allow information to persist;
the effect is often described as memory.
Recurrent neural networks

 RNN use cases tend to be connected to language models in

which knowing the next letter in a word or the next word in a
sentence is predicated on the data that comes before it. A
compelling experiment involves an RNN trained with the works
of Shakespeare to produce Shakespeare-like prose --
successfully. Writing by RNNs is a form of computational
creativity. This simulation of human creativity is made possible
by the AI’s understanding of grammar and semantics learned
from its training set.
Applications of Recurrent Neural Networks
https://vinodsblog.com/2019/01/07/deep-learning-introduction-to-recurrent-
neural-networks/
Applications of Recurrent Neural Networks
RNNs
Applications of Recurrent Neural Networks
RNNs support processing of sequential data by the addition of a loop. This loop
allows the network to step through sequential input data whilst persisting
the state of nodes in the Hidden Layer between steps - a sort of working
memory. The following image gives a conceptual representation of how this
works. A RNN can be viewed as many copies of a Feed Forward ANN
executing in a chain.
Applications of Recurrent Neural Networks
RNNs are widely used in the following domains/ applications:
 Prediction problems
 Language Modelling and Generating Text
 Machine Translation
 Speech Recognition
 Generating Image Descriptions
 Video Tagging
 Text Summarization
 Call Center Analysis
 Face detection, OCR Applications as Image Recognition
 Other applications like Music composition
 Welcome
Thank you so much for your attention in
this webinar. Other information about each
section and topic can be shared and sent
to you by having this initial background
about Machine Learning and Artificial
Intelligence.

View publication stats