Machine Learning










PDF
version of this report

You must have Adobe Acrobat reader to view, save, or print PDF files. The
reader is available for free
download
.

Machine Learning

by James
G. Barr

Docid: 00021056

Publication Date: 2207

Report Type: TUTORIAL

Preview

Machine learning (ML) is an artificial intelligence (AI) technology that
enables a system to enhance its awareness and capabilities – that is, to
learn – without being explicitly programmed to do so. In some cases, ML
systems learn by studying information contained in data warehouses. In
other cases, ML systems learn by conducting thousands of data simulations,
detecting patterns, and drawing inferences.

Report Contents:

Executive Summary

[return to top of this report]

Machine learning (ML) is an artificial intelligence (AI) technology that
enables a system to enhance its awareness and capabilities – that is, to
learn – without being explicitly programmed to do so.

Artificial Intelligence Tutorial
Smart Machines Tutorial

In some cases, ML systems learn by studying information contained in data
warehouses. In other cases, ML systems learn by conducting thousands of
data simulations, detecting patterns, and drawing inferences.

ML systems don’t deduce the truth as humans do; rather they forecast a
perceived
truth based on available data. As analyst Nick Heath observes, “At a very
high level, machine learning is the process of teaching a computer system
how to make accurate predictions when fed data. Those predictions could
be:

  • “Answering whether a piece of fruit in a photo is a banana or an
    apple,
  • “Spotting people crossing the road in front of a self-driving car,
  • “[Determining] whether the use of the word ‘book’ in a sentence relates
    to a paperback or a hotel reservation,
  • “[Deciding] whether an e-mail is spam, or
  • “Recognizing speech accurately enough to generate captions for a
    YouTube video.”1

Origins of ML

The term “machine learning” was first defined by Arthur Samuel in 1959 as
“the ability to learn without being explicitly programmed.”2
The concept gained traction in the 1990s when advances in storage and
processing technology (specifically, massively parallel processing)
enabled enterprises to create and curate enormous data warehouses where
financial and customer transactions, as well as other information, could
be stored and analyzed in an effort to identify and exploit
previously-unknown competitive advantages. In this context, machine
learning is simply the latest technological tool for performing
large-volume data analysis – both for structured and, increasingly,
unstructured data.

Machine Learning Adoption

Analyst Jay Selig reports that machine learning is rapidly gaining
enterprise acceptance:

  • “In 2021, 41 percent of companies accelerated their rollout of AI as a
    result of the pandemic.
  • “These newcomers are joining the 31 percent of companies that already
    have AI in production or are actively piloting AI technologies.”3

Machine Learning Growth

The application of machine learning to enterprise operations is expected
to expand, even accelerate, owing to:

  1. The exponential growth in big data; in particular, data produced by
    Internet of Things (IoT) platforms, systems, applications, and sensors.
  2. The increasing generation of “synthetic” data through data
    extrapolation and simulation.
  3. The steady advancements in machine learning algorithms, making
    machines smarter.
  4. The collapsing costs of storage infrastructure, making ML affordable.
  5. The transformative effects of machine learning on business processes,
    helping enterprises realize the goals of 90s-era business process
    reengineering theory.
  6. The influence of machine learning on robotics and other allied AI
    fields.
  7. The ability to displace expensive blue- and white-collar personnel, as
    executives recognize that technology, not globalization, is the real
    engine of enterprise cost-cutting and profitability.

Description

[return to top of this report]

A computer program is said
to learn from experience E with respect to some task T
and performance measure P if its performance on T,
as measured by P, improves with experience E.

As an area of keen interest to both the private and public sectors, a
report from President Obama’s Executive Office of the President: National
Science and Technology Council Committee on Technology saw machine
learning as “a statistical process that starts with a body of data and
tries to derive a rule or procedure that explains the data or can predict
future data.

“This approach – learning from data – contrasts with the older ‘expert
system’ approach to AI, in which programmers sit down with human domain
experts to learn the rules and criteria used to make decisions, and
translate those rules into software code. An expert system aims to emulate
the principles used by human experts, whereas machine learning relies on
statistical methods to find a decision procedure that works well in
practice.”4

As described by analyst Singh Niven, “Developing machine learning
applications is different than developing standard applications. Instead
of writing code that solves a specific problem, machine learning
developers create algorithms that are able to take in data and then build
their own logic based on that data.”5

ML Algorithms

Machine learning methods rely on a wide variety of complex algorithms,
including:

  • Decision trees
  • Ordinary least squares regression
  • Clustering

Clustering, for example, is the process of grouping a set of objects such
that objects in the same group (or cluster) are more similar to each other
than to objects in other groups. Perhaps most familiar to consumers, ML
algorithms allow Netflix to make movie recommendations based on
recently-viewed films and permit Amazon to predict what new books a person
might prefer based on past purchases.6

While the algorithmic approach to machine learning is generally
productive, analyst Daniel Faggella reports that “researchers have found
that some of the most interesting questions arise out of none of the
available machine learning algorithms performing to par.” Although this
phenomenon may be attributed most often to bad training data, it also
occurs when working with new domains.7

Conversational Commerce

One prominent product made possible by machine learning is the
“intelligent personal assistant” (IPA) – also known as a “smart personal
assistant,” “intelligent virtual assistant,” “virtual digital assistant,”
or simply “virtual assistant.” The IPA is essentially a software program
that helps people complete basic tasks. Typically, an IPA will answer
questions and perform actions based on voice commands and location
awareness. Popular models include Apple’s Siri and Amazon’s Alexa.
Responding instantly to user requests, which include searching,
purchasing, controlling connected devices, and “facilitating professional
tasks and interactions,” analyst Joanna Goodman sees IPAa ushering in an
era of “conversational commerce,” offering:

  • Instant consumer gratification
  • Instant revenue for businesses

As employed by enterprise personnel, IPAs provide instant responses to
queries, thus improving productivity and job satisfaction.8

Deep Learning

As evaluated by President Obama’s Executive Office of the President:
National Science and Technology Council Committee on Technology, “some of
the most impressive advancements in machine learning have been in the
subfield of deep learning, also known as deep network learning.

“Deep learning uses structures loosely inspired by the human brain,
consisting of a set of units (or ‘neurons’). Each unit combines a set of
input values to produce an output value, which in turn is passed on to
other neurons downstream. For example, in an image recognition
application, a first layer of units might combine the raw data of the
image to recognize simple patterns in the image; a second layer of units
might combine the results of the first layer to recognize
patterns-of-patterns; a third layer might combine the results of the
second layer; and so on.”9

As illustrated in Figure 1, deep learning (DL) is a logical subset of
machine learning (ML), with machine learning being a subset of artificial
intelligence (AI).

Figure 1. Artificial Intelligence – Machine Learning – Deep Learning

Figure 1. Artificial Intelligence - Machine Learning - Deep Learning

Source: Wikimedia Commons

Forms of Machine/Deep Learning

According to analyst Karen Hao, both machine and deep learning
methodologies come in three basic “flavors”:

  1. Supervised Learning – “The most prevalent [variety],
    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.”
  2. 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 learning has “gained traction
    in cybersecurity.”
  3. Reinforcement – The newest methodology, “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.”10

A fourth form of machine/deep learning is called
Semi-supervised Learning
, in which algorithms train on small
sets of labeled data and apply what they discern to unlabeled data. This
approach, it must be observed, is often invoked in the absence of quality
data.11

Learning Techniques

Just as humans employ multiple learning techniques to increase their
knowledge and inform their decision-making, machine learning practitioners
have devised multiple, scenario-based approaches to artificial learning.
As itemized by analyst Jason Brownlee, these techniques include:

Multi-Task Learning – “A type of supervised learning
that involves fitting a model on one dataset that addresses multiple
related problems.”

Active Learning – “A technique where the model is able
to query a human user operator during the learning process in order to
resolve ambiguity during the learning process.”

Online Learning – A technique which “involves using the
data available and updating the model directly before a prediction is
required, or after the last observation was made.”

Transfer Learning – “A type of learning where a model is
first trained on one task, then some or all of the model is used as the
starting point for a related task. It is different from multi-task
learning as the tasks are learned sequentially … , whereas multi-task
learning [is applied to multiple tasks] in parallel.”

Ensemble Learning – “An approach where two or more
[models] are fit on the same data and the predictions from each model are
combined.”12

These learning techniques support and reinforce common business values,
such as:

  • Promoting resource “reusability” – Multi-Task Learning and Transfer
    Learning
  • Encouraging mentoring (in this case, human-to-machine) – Active
    Learning
  • Maintaining process relevance and viability (via timely business
    process reengineering) – Online Learning
  • Validating and refining operations – Ensemble Learning

Applications

[return to top of this report]

Machine learning forms the foundation of a broad range of enterprise
applications, as itemized by Tata Consultancy Services in Figure 2.

Figure 2. Machine Learning Applications By Industry Sector

Figure 2. Machine Learning Applications By Industry Sector

Source: Tata Consultancy Services

Cancer Detection

One area where machine learning shows huge promise is detecting cancer in
computer tomography (CT) imaging. First, researchers assemble as many CT
images as possible to use as training data. Some of these images show
tissue with cancerous cells, and some show healthy tissues. Researchers
also assemble information on what to look for in an image to identify
cancer. For example, this might include what the boundaries of cancerous
tumors look like. Next, they create rules on the relationship between data
in the images and what doctors know about identifying cancer. Then they
give these rules and the training data to the machine learning system. The
system uses the rules and the training data to teach itself how to
recognize cancerous tissue. Finally, the system gets a new patient’s CT
images. Using what it has learned, the system decides which images show
signs of cancer, faster than any human could. Doctors could use the
system’s predictions to aid in the decision about whether a patient has
cancer and how to treat it.13

Robot Learning

Machine learning is becoming a key enabling technology of robot learning,
with core research being conducted at major institutions like the
Massachusetts Institute of Technology (MIT) and Tokyo University, as well
as high-tech innovators like Google.14

As analyst Karen Hao explains, “Existing reinforcement-learning
algorithms that allow robots to learn movements through trial and error
still rely heavily on human intervention. Every time the robot falls down
or walks out of its training environment, it needs someone to pick it up
and set it back to the right position.

“Now a new study from researchers at Google has made an important
advancement toward robots that can learn to navigate without this help.
Within a few hours, relying purely on tweaks to current state-of-the-art
algorithms, they successfully got a four-legged robot to learn to walk
forward and backward, and turn left and right, completely on its own.”15

Physical Sciences

Researchers report that “In parallel to the rise of ML techniques in
industrial applications, scientists have increasingly become interested in
the potential of ML for fundamental research, [including physics]. To some
extent, this is not too surprising, since both ML and physics share some
of their methods as well as goals. The two disciplines are both concerned
about the process of gathering and analyzing data to design models that
can predict the behavior of complex systems. However, the fields
prominently differ in the way their fundamental goals are realized. On the
one hand, physicists want to understand the mechanisms of Nature, and are
proud of using their own knowledge, intelligence and intuition to inform
their models. On the other hand, machine learning mostly does the
opposite: models are agnostic and the machine provides the ’intelligence’
by extracting it from data. Although often powerful, the resulting models
are … as opaque to our understanding as the data patterns themselves.
Machine learning tools in physics are therefore welcomed enthusiastically
by some, while being eyed with suspicions by others. What is difficult to
deny is that they produce surprisingly good results in some cases.”16

Development Services

To help facilitate the development of machine learning applications,
leading platform-as-a-service (PaaS) providers offer machine learning
services, such as:

  • Amazon SageMaker
  • Azure Machine Learning
  • Google Vertex AI
  • (IBM) Watson Machine Learning

Challenges

[return to top of this report]

“Machine learning can’t
get something from nothing … what it does is get more from less.”

– Dr. Pedro
Domingo, University of Washington17

Despite the tremendous potential for turning raw data into actionable
intelligence, machine learning has some controversial elements.

Non Transparency

Test-taking students are frequently admonished to “show your work,” as a
way of proving that their answers were formulated through logical means.
Unfortunately, one of the troubling aspects of machine learning – and AI
in general – is often the inability to conduct human oversight,
particularly as the process of ML decision-making becomes more
sophisticated, and less grounded in traditional data analysis techniques.
The rationale for ML conclusions may, in some cases, be unexplained and
unexplainable, creating a potential crisis in confidence.

Data Security

As with other information technologies, machine learning systems are
vulnerable to attack. Exposures exist on two levels:

  • First, compromised data could result in ML applications “learning the
    wrong lessons.”
  • Second, compromised applications could result in erroneous data
    interpretations.

Either condition could adversely affect enterprise operations.

Adversarial Machine Learning

Adversarial machine learning (AML) is the term applied to efforts aimed
at fooling, or otherwise disrupting, machine learning models. Common
tactics include:

  • Presenting a model with inaccurate data during training; and
  • Offering “maliciously designed” data to an already trained model.18

More broadly, AML is concerned with the design of ML algorithms that can
resist security challenges, the study of the capabilities of attackers, and
the understanding of attack consequences.

To help facilitate AML development, the US National Institute of
Standards and Technology (NIST) has advanced “A Taxonomy and Terminology
of Adversarial Machine Learning.” The report is intended as a step toward
securing AI applications, especially against adversarial manipulations of
machine learning, by developing an AML taxonomy and terminology.19

Employment Anxiety

Machine learning, like other forms of artificial intelligence, will
disrupt job markets in unanticipated ways. According to the McKinsey
Global Institute, “Dealing with job displacement, retraining, and
unemployment will require a complex interplay of government, private
sector, and educational and training institutions, and it will be a
significant debate and an ongoing challenge across society.”20

Uncontrolled Evolution

Machine learning is evolving at an exciting – and, to some, an alarming –
rate. Even former US Secretary of State Henry Kissinger is asking
questions. “The implications of this evolution are shown by a recently
designed program, AlphaZero, which plays chess at a level superior to
chess masters and in a style not previously seen in chess history. On its
own, in just a few hours of self-play, it achieved a level of skill that
took human beings 1,500 years to attain.

“If AlphaZero was able to achieve this mastery so rapidly:

  • “Where will AI be in five years?
  • “What will be the impact on human cognition generally?
  • “What is the role of ethics in this process?”21

Recommendations

[return to top of this report]

“Machine-learning
algorithms find and apply patterns in data. And they pretty much run the
world.”

– Karen Hao22

Implementing Machine Learning

While implementing machine learning is seldom fast and easy, the
following six-step process, as offered by analyst Ed Burns, offers a
relatively straightforward process:

  1. Identify a Problem – “The most effective machine
    learning projects tackle specific, clearly defined business challenges
    or opportunities.”
  2. Choose an Algorithm – “Different machine learning
    algorithms are better suited for different tasks. Cutting-edge deep
    learning algorithms are better at complicated things like image
    recognition or text generation.”
  3. Gather Relevant Data – “Data collection involves
    complicated tasks like identifying data stores, writing scripts to
    connect databases to machine learning applications, verifying data,
    cleaning and labeling data and organizing it in files for the algorithm
    to work on.”
  4. Build the Model – “This step … will differ
    substantially depending on whether the [ML] team is using a supervised
    machine learning algorithm or an unsupervised algorithm.”
  5. Develop the Application – “Now that the algorithm
    has developed a model of what the data looks like, data scientists and
    developers can build that learning into an application that addresses
    the business challenge or opportunity identified in the first step of
    the process.”
  6. Validate the Model – “Data scientists should verify
    that [the] application is delivering accurate predictions on an ongoing
    basis.”23

Business Process Reengineering

The goal of every enterprise is continuous improvement, a key feature of
machine learning.

Analyst Shuvro Sarkar advises that machine learning is ideal for
business process reengineering since “[ML] algorithms are iterative in
nature, repeatedly learning and probing to optimize outcomes. Every time
an error is made, machine learning algorithms correct [themselves] and
[begin] another iteration of the analysis. And all of these calculations
happen in milliseconds making it exceptionally efficient at optimizing
decisions and predicting outcomes.”24

Personnel Planning

Machine learning is a disruptive technology which enables information
systems to perform functions previously performed by humans. As machine
learning pervades the enterprise space, workforce reductions are
inevitable.

As jobs disappear, the ethical enterprise has an obligation to mitigate
the impact on employees and their families by:

  • Retraining valued workers, permitting them to assume new roles within
    the enterprise; or
  • Providing outplacement services, positioning displaced workers to find
    gainful employment elsewhere.

These strategies should be codified in a personnel plan designed
specifically to manage machine learning-related human resources issues.

References

[return to top of this report]

[return to top of this report]

About the Author

[return to top of this report]

James G. Barr is a leading business continuity analyst
and business writer with more than 30 years’ IT experience. A member of
“Who’s Who in Finance and Industry,” Mr. Barr has designed, developed, and
deployed business continuity plans for a number of Fortune 500 firms. He
is the author of several books, including How to Succeed in Business
BY Really Trying
, a member of Faulkner’s Advisory Panel, and a
senior editor for Faulkner’s Security Management Practices. Mr. Barr can
be reached via e-mail at jgbarr@faulkner.com.

[return to top of this report]