Machine learning and artificial intelligence are frequently used interchangeably.
As a result, they must interact with one another.
But how do you do it?
Artificial intelligence, or AI, is the study of teaching robots to do human- like activities.
When scientists began investigating how computers might solve problems on their own in the 1950s, they coined the term.
When you hear about AI, you're probably thinking about these absurd visuals.
Ravenna, Andy:
I grew up watching "The Jetsons" and the movie "2001: A Space Odyssey" both of which featured HAL the talking computer.
Artificial intelligence, in my opinion, is a computer or a system that has been given human- like capabilities.
Every second of every day, we take for granted how our brains automatically compute the environment around us.
AI refers to the idea that a machine can do the same tasks as a human.
Rogers, Bob:
All of our gadgets' voice interfaces are artificial intelligence (AI).
And it's fantastic.
You are allowed to use an accent.
You can speak a certain dialect, and AI systems can quickly learn how to engage with you if there is material on the Internet in that dialect.
I can use my phone to ask it questions, such as "Siri, say my name."
SIRI: Your name is Tom.
But because we're friends, I'm allowed to refer to you as M.
Smiley, R.
Computer learning is a subset of AI that teaches a machine how to learn. While AI is the broad science of replicating human skills, machine learning is a specialized subset of AI that teaches a machine how to learn.
Machine learning algorithms, like you and me, seek for patterns in data and try to make conclusions.
Rogers, Bob:
People are not directly programming them.
You can really offer them some instances, and they'll pick up on what you're saying.
That's a significant distinction since it's far easier for us to offer examples than it is to write code.
When the algorithm has mastered generating the correct conclusions, it applies what it has learned to fresh sets of data.
Rogers, Bob:
That's the life cycle in a nutshell.
It is to ask the question, gather the data, train the algorithm, test it, collect feedback, and utilize the feedback to improve the algorithm so that accuracy and performance improve.
It's done!
Machine Learning is a term that refers to the study of
Lisa Loftis (Loftis):
When you look at the Google car, you'll notice that it has lasers on the top that tell it where it is in relation to the rest of the world.
It has radar in front of it that tells it the speed and movements of all the automobiles surrounding it.
And it utilizes all of that information not only to figure out how to drive the car, but also to forecast what future drivers in the area will do.
And that automobile is processing roughly a gigabit of data per second.
Ravenna, Andy:
For instance, one of the projects they're working on is tumor scanning.
When they scan the outside of the body using X-
rays and photos, they are hunting for tumors.
And can you image a day when instead of having to go in and take a sample every time they want to find out what's going on inside your body, they could just perform a scan to see if a tumor is benign or not?
Isn't it extremely Star Trek?
Lisa Loftis (Loftis):
And, when you consider the quantity of constantly streaming data from IoT beacons and sensors, it allows us to comprehend our surroundings far more deeply than we did previously.
That's exactly what AI and machine learning require.
They require data at the granular level.
They require extremely vast amounts of data.
They also require a wide range of data sources in order to discover trends and learn.
BORNE, KIRK:
So, if we can have a chat bot inform us the newest news in the morning, that would be great.
So it can read the stock market, right?
It can read all of the financial statistics we require and, in a human-
interpretable manner, tell you not only the numbers but also the story behind the numbers.
That, I believe, is quite valuable.
When I think about what's fresh and coming up, I'm not sure if mathematics or the sorts of activities we do are changing as much.
That is all there is to it.
Machine learning is the mechanism through which machines learn from data, while AI is the study of computers simulating humans.
There are a plethora of data-
driven challenges to tackle.
Why not delegate a few to an algorithm? 
Today, I'm going to talk to you about the many sorts of machine learning problems you're likely to run across while implementing AI systems at your organization.
Classification (both binary and multi-
class regression), which is a continuous variable or frequently a numerical outcome, time series, where we want to forecast over a specified time window, and anomaly detection, where we're looking for outliers in our data, are the forms I'm going to discuss today.
It's a subset of classification for binary classification.
There are just two potential classes in binary categorization.
This might be a yes/no or true/false question.
When it comes to loan underwriting, a bank will employ binary categorization.
The debt may only fall into one of two categories or outcomes: yes, it will default, or no, it will not default.
A bank is interested in learning about a loan's default rates in order to prevent approving problematic loans.
A subset of classification issues is multi-
classification. It, however, differs from binary classification in that it contains more than two potential results.
This might be a yes, no, or possibly answer.
It may be red, yellow, or blue in color.
A, B, C, or D are all possibilities.
Let's take a look at how a telecoms business may deal with a multi-
class issue. The telecom gets a list of customers who are most likely to churn or terminate their contracts every month.
As a result, the telecom comes up with four potential offers that it is ready to make in order to retain its customers.
A discount on their existing plan, a phone upgrade, a plan switch, or even extra data are all possibilities.
They'll utilize multi-
classification to estimate which of these four offers each subscriber will respond to the most.
The telecom retains its income and enhances the lifetime value of its subscribers by reducing subscriber churn.
Then there's regression.
We're predicting a continuous variable, which is usually a numerical number, with regression.
Regression might be used by a hospital to forecast how long a patient will stay at their facility once accepted.
If the patient is just going to be there for one day, that's a lot different than if they are going to be there for 14 or even 30 days.
Understanding how long their patients are expected to stay using regression may help them arrange supplies, personnel, and other resources needed to care for them.
The time series comes next.
We're predicting across a specific time span with time data.
This may happen in the next six hours, seven weeks, or perhaps eight years.
Now that I operate a clothes firm, I need to know how much inventory I'll need in the coming quarter.
To forecast my daily sales for the following three months, I'd utilize time series.
Then, armed with that knowledge, I can guarantee that I make orders to ensure that I have enough inventory to meet those projected sales.
This is crucial to my business's operations because if I order too much inventory, I'll end up wasting money on items I'll never sell.
Customers will not come to my business if I buy too little inventory because they will not be able to get their orders filled.
Finally, we'll discuss anomaly detection.
This is known as "outlier detection," and it's used to locate abnormalities or outliers in your data.
This may be used for a variety of purposes, including identifying fraud, forecasting hardware failure, and detecting human mistake in data.
Consider the IT department of a firm.
They'd be quite interested in employing anomaly detection to protect themselves against cyber-
attacks.
Consider
the case of a corporation that is entirely located in the United States.
As a result, the vast majority of their network traffic should originate in the United States.
However, if we notice network activity coming from another nation, this should certainly arouse worry and prompt additional inquiry.
Anomaly detection can notice those anomalies and strange activities for us, triage it, and report it to the IT department, where it can be investigated quickly to avert security issues.
So those are the four most prevalent machine learning issues.
It's critical to know which problem you're working on since various problems require different models, accuracy measurements, and other problem-
specific aspects that must be taken into account. 
No comments:
Post a Comment
If You Have Any Doubts, please Let Me Know