Professional Documents
Culture Documents
Analysis and Design of Algorithm (3150703) : Assignment-1
Analysis and Design of Algorithm (3150703) : Assignment-1
Assignment-1
5th -IT
Machine Learning is the ability to accurately perform new, unseen tasks, built
on known properties learned from training or historic data, and based on
prediction. Or in simpler terms, machine learning is when algorithms parse
data, learn from that data, and then apply their learnings to make informed
decisions.
You can think of it like a computer that can essentially teach itself to
continually improve on performing an automated task based on analyzing the
results of past attempts. Given that computers can process huge amounts of
information at high speed, it can thus learn very fast.
To illustrate a simple example of machine learning in action, let’s go back to
the birth of machine learning. In 1956, when Arthur Lee Samuel didn’t want to
write a highly-detailed, lengthy computer program that could beat him in a
game of checkers, he created an algorithm instead.
This algorithm allowed the computer to play against itself over and over
again. The algorithm could then learn from the historic data and essentially
teach itself to perform against an opponent. The effectiveness of the
algorithm was proven when, in 1962, the computer beat the Connecticut state
champion in a game of checkers.
Deep Learning on the other hand is a new and exciting subset of Machine
Learning based on neural net technology. Through a complex set of
algorithms, it allows a machine to discover new patterns without being
exposed to any historical or training data.
Leading startups in this space are DeepMind, bought by Google in early 2014
for $500 million, back when DeepMind had just 13 employees. Another
leading startup in this space is Vicarious, funded with investment from Elon
Musk, Jeff Bezos and Mark Zuckerberg.
Twitter, Baidu, Microsoft, and Facebook also heavily invest in this area. Deep
Learning algorithms rely on discovery and self-indexing, and operate in much
the same way that a baby learns first sounds, and then words, sentences and
even languages.
Most companies recognize that the explosion of data from sensors and
intelligent devices and new information from social media and other sources
make now the right time to invest in a more comprehensive algorithm
strategy. The truth is that most companies have a hard time managing the
streams of data that they already have, let alone handling new sources of data.
Algorithms are the key to making your current data useful and making smart,
timely decisions based on that data possible.
What has been missing thus far is focus on how to manage this critical resource and
the assets they produce. Markets will be won by companies that effectively create,
manage, and deploy algorithms for everything from customer demand sensing to
data cleansing to inventory management! Companies that are faster and smarter at
developing and managing algorithms are going to win in industry after industry.
And yet, almost all companies report difficulty in moving ahead for several reasons:
No one owns algorithms. They exist across business units and functions.
No one is setting priorities and deciding where algorithms can make
customers the happiest. Few are aligning technology investments
against the most important algorithms.
Algorithms exist in every company. They might not be backed by
accurate data. They might not be shared between business units. But
they exist and are being used to influence or drive some decisions.
People know that a more data-based, analytically driven, algorithm
supported management process is required. They just want
management to be smart about setting priorities and getting algorithms
better every day.
It is amazing how much better decisions are when, for example, an algorithm
has analyzed the past eight years of purchasing data, ingested the latest social
media buzz, and considered customer usage patterns as reported back by a
sensor built into the product. Even the most experienced human demand
planner with an Excel tool (with macros) can’t compete!
Bibliography