Two Minute Technology: Machine Learning

Two Minute Technology: Machine Learning

Any learning entity - be it an animal, human or a machine begins with an initial set of observations - raw data. This kind of data can stem from interactions, transactions, information exchange, examples, instructions or experiences. A brain - whether it belongs to a human or animal, tries to look for patterns inside that initial data and then uses it to perform actions or make decisions. This same phenomenon can be applied to machines, using machine learning.

Machine learning is a subfield of Artificial Intelligence. According to Arthur Samuel, a computer scientist, machine learning gives computers the ability to learn without being explicitly programmed. Another computer scientist, Tom Mitchell, took the same details of machine learning, but broke it down further in his definition. I find Mitchell's definition more fitting to describe exactly what machine learning is.

"Machine learning is a well-posed learning problem: a computer program is said to learn from experience (E) with respect to some task (T) and some performance measure (P), if its performance on T, as measured by P, improves with experience E." - Tom Mitchell (1998)

Suppose your email program watches which emails you do or do not mark as spam, and based on that, this is how the computer learns how to better filter spam. Using Mitchell's definition of machine learning, what is the task (T) in this scenario?

  • Watching the user label emails as spam or not spam would be experience (E)
  • The number (or fraction) of emails correctly classified as spam or not spam would be performance (P)
  • So, the act of classifying emails as spam or not spam would be the task (T)
Image from https://www.analytixlabs.co.in/blog/machine-learning-topics/#sub1

The above figure details the process flow or life cycle of machine learning. Where the core four processes during the machine learning life cycle are:

  1. Gathering data (input)
  2. Analyzing the data
  3. Selecting the best algorithm to organize the data
  4. Data visualization/implementation (output)

What Are the Topics in Machine Learning?

Algorithms of Machine Learning

This section will cover the very basics of the most common types of problems that machine learning can be used to solve, and two common learning algorithms.

  • Supervised: teaching the machine how to do something.
    • We give the computer data that already has the correct relationship between input and output.
    • Can be used to solve regression and classification problems
      • Regression: predict continuous valued output (i.e. price, age)
      • Classification: discrete valued output (i.e. 0 or 1, fresh or rotten)
  • Unsupervised: letting the machine learn by itself.
    • We can derive structure from unlabeled data, without necessarily knowing the effect of the variables.
    • Can be used to solve clustering problems
      • Clustering: classifying data into natural groups based on similarities (i.e. demographics, interests)

Message from the Blogger

Hello! Thank you for reading this blog post. I hope you enjoyed it and learned something new today. See you next time!

Comments

Post a Comment

Popular posts from this blog

How apple has dominated the phone company

Image
  In the growth of smartphones in 2007  apple made basically the first touch screen held cell phone in which it was different from all of the others that was around. This tech guru Steve jobs knew at the time that he had something special. At this time he was competing with companies that were making flip phones, blackberries or even the slide up phones at the time. This was an outside the box move in which no one was doing at the time or even thought about doing.  Cell phones have come a long way back in the day before that they were carrying like bricks with them and you can tell it was a bad time. This mix of new styles of phones which basically turned the whole industry into touch screens and smartphones. This company lost their lead Steve jobs who was the man with the plan and stated this trend but the company has been able to continue to grow.  Many people don't know that the company made a bad deal with AT&T at the time. This deal was made to get all of t...
Read more

Ryan's Top Ten Albums - In Rainbows - Radiohead

Image
  In Rainbows - Radiohead Released 2007 - Genre: Art Rock, Experimental Rock While the albums I talked about in this blog are not ranked in any particular order, I can confidently say that this is my favorite album. Radiohead has always been one of my favorite bands, my original idea for this blog was to do an entry on each of their nine studio albums. In Rainbows came at a very unique point in the band’s career, as they had already seemingly reinvented the wheel twice, with the release of OK Computer in 1997, and Kid A in 2000, and their few albums since had been highly regarded but never to that same level. In Rainbows was their attempt at reinventing their sound again and doing something groundbreaking, and they delivered. The buildup to the album's release was very unique, as the band promoted the album almost solely on their own website, where they boasted a pay-what-you-want scheme for the album, where fans could pay any amount (or nothing at all) to download the album. I...
Read more