### Introduction

Machine learning works on the development of computer programs that can access data and use it to automatically learn and improve from experience. You can watch a robot builder construct a house in just two days. This was back in July 29th, 2016, so that's pretty impressive, this amount of time to continue to grow in his development. And it's smart enough to leave spaces in the brickwork for wiring and plumbing and can even cut and shape bricks to size.

Amazon Echo relies on machine learning and with more data it becomes more accurate. Play your favorite music, order pizza from dominoes, voice control your home, request rides from Uber.

Have you ever wondered the artificial intelligence, a technique which enables machines to mimic human behavior? This is really important because this is how we are able to gauge how well our computations or what we're working on works is the fact that we're mimicking human behavior. We're using this to replace human work and make it more efficient and make it more streamlined and more accurate. And so the center of artificial intelligence is the big picture of all this put together.

IBM deep blue chess, electronic game characters. Those are just a couple examples of artificial intelligence.

### What is Machine Learning?

Machine learning, is a technique which uses statistical methods enabling machines to learn from their past data. So this means if you have your input from last time and you have your answer, you use that to help prove the next guess it makes for the correct answer. IBM Watson, Google search algorithm, email spam filters, these are all part of machine learning.

And then deep learning, which is a subset of machine learning, composing algorithms that allow a model to train itself and perform tasks. AlphaGo, natural speech recognition. These are a couple examples.

Deep learning is associated with tools like neural networks where it's kind of a black box. As it learns, it changes all these things that are, as a human, we'd have a very hard time tracking. That's able to come up with an answer from that.

### How Machine Learning Works

Now let's see how machine learning works. First we start with training the data. Once we've trained the data, we go into the machine learning algorithm, which then puts the data into a processing, which then goes down to another machine learning algorithm. And then we take new data, because you have to test whatever you did and make sure it works correctly, and we put that into the same algorithm.

Once we do that, we check our prediction, we check our results, and from the prediction, if we've set aside some training data and we find out it didn't do a good job predicting it, and it gets a thumbs down, as you see. Then we go back to the beginning and we retrain the algorithm. And a lot of times it's not just about getting the wrong answer, it's about continually trying to get a better answer.

So you'll see the first time you might be like, oh, this is not the answer I want, depending on what domain you're working in, whether it's medical, economical, business, stocks, whatever. You try out your model, and if it's not giving you a good answer, you retrain it. If you think you can get a better answer, you retrain it. And you keep doing that until you get the best answer you can.

### Types of Machine Learning

- Supervised learning and
- Unsupervised learning.

#### Supervised Learning

This is where we deal with a known amount of data, like for example we have a bunch of apples and we have a machine learning algorithm. It goes through the process. It goes through and trains a model based on that known data. And then once you've trained your model on the known stuff, you can then put an unknown data in there, and you get a new response. And of course in this particular one, it's an apple. So it's trying to figure out whether it's an apple or another fruit.

There are many different algorithms you can use for computing this information, for doing this supervised training. Just to list some of the top ones that are currently being used, and by no means there's more than just this. So by no means this isn't the complete list.

- There's polynomial regression.
- There is a random forest.
- There's linear regression.
- There's logistic regression.
- There's decision trees.
- There's k-nearest neighbors.
- And there's naive Bayes.

Like I said, this is just a short list of some of the many tools that are out there nowadays.

#### Unsupervised Learning

So, if you have supervised learning, then we should also look at unsupervised learning. Unsupervised learning, in this case we will deal with unknown data. For example we can have a bunch of frui and we might not have labeled it, we don't know, we've never had anybody look at it and say, this is what this is. And we take that data and we put it through the machine learning algorithm. And then that goes through the processing and then the trained model. And what the trained model says, hey, can I see a pattern here? And from that pattern it divides it up into a response, in this case, apples and pears.

As we know, some of these things look just like the other and it tries to put them all together so that you get similar things in similar groups.

So again, as an example in unsupervised learning, we have a nice list of algorithms here and this is not the only algorithms used for this, so don't limit yourself to just these. These are just some of the primary ones used today. And of course we have:

- K-means clustering
- Singular value decomposition
- Fuzzy means
- Partial least squares
- Apriori
- Hierarchical clustering
- Principal component analysis.

### Machine Learning Prerequisite

Machine learning prerequisites, computer science, fundamentals and programming. So any of the machine learning out today, you have to know some basic scripting or programming.

Intermediate statistical knowledge. You have to understand a little bit about probabilities. If A is current, how likely is B going to happen? If there's clouds overhead, how likely is it going to rain?

Linear algebra and intermediate calculus. The linear algebra is very important because you have to understand basically drawing a line through the data points and what that means. That's the most fundamental linear regression model is you draw a line through all your data and you use that line to compute new values.

Intermediate calculus means you need to have a little bit of understanding of what a differential equation is. You really don't need to be an expert because the computer does all the heavy lifting for you, but it's important to know the terms when they come up unless you're doing some advanced programming on the actual models themselves.

Then, data wrangling and cleaning. I would say this might be the biggest one in here you have to start getting a grip on how to clean up your data. There's a saying that bad data in, bad data out. So when good data is in, you're more likely to have good data out.

### Application of Machine Learning

Some applications of machine learning includes:

- Instant segmentation
- Object detection

### Conclusion

So to summarise it up, we covered the basics of machine learning. What is machine learning? We talked a little bit about the process or the workflow of machine learning. We've looked at two different divisions of machine learning, supervised and unsupervised. We went over the prerequisites you should have going into machine learning. Then you should have the basic fundamentals or a little bit of computer science and programming or scripting skills.

If you enjoyed this post please be kind enough to share it with your friends and don't forget to comment any of your doubt in the comment section below. We will answer you as soon as possible.

Print this post
## 0 Comments