Source: Deep Learning on Medium
Introduction to Machine Learning for the Non-Techie Person
Machine Learning (ML) has been a buzz word for several years now. It has actually been around for quite a while — decades even. It’s a subfield of AI or artificial intelligence.
No, we’re not talking about cyberpunk theories where machines will become self-aware and take over the earth. What we actually have are programs and devices that can actually help human beings live life better.
Uses of Machine Learning
Today machine learning is used in a wide range of applications. They are used in industries anywhere from automating mundane tasks all the way to providing intelligent insights on market data.
Don’t be surprised to find out that you may already be using devices and applications equipped with machine learning. If you own a smartphone or maybe you’re using web services then you have already encountered machine learning first hand.
Here are some of the common uses of this technology today:
- Image Recognition — have you ever used QR codes? Or how about facial recognition to unlock your phone — phones have that feature, right?
Anything from optical character recognition to accessing and using photos of people on a database is part of image recognition.
- Speech Recognition — you may have heard of Alexa, the voice-powered digital assistant developed by Amazon for their Echo devices and speakers. Smartphones today can also do voice searches — well voice searching has been provided by Google for quite a while now.
Any technology that involves the translation of spoken words into text makes use of speech recognition. Voice commands on smart devices like your TV, locks, lights, or any other appliances also make use of machine learning.
- Medical Diagnosis — machine learning is also a technology being used by doctors for medical diagnosis. This tech is even used to detect cancerous tissue cells.
- Financial Trading — machine learning is a technology that is also used to predict market trends, profitability, and also in fraud detection.
Why We Need Machine Learning Today
It is no secret that data from many different sources is growing exponentially today. We have customer data, web service user data, statistical data, and a whole lot of others.
Note that the majority of data that we have gathered through the decades is unstructured. Around 80% of it is just a huge chunk of graphs, documents, photos, videos, audio, and many others.
Understanding that the amount of data is impossible by human ability alone. Finding and patterns in that data to make it useful for us is also impossible. The data is just too massive.
And this is where machine learning has been put to its best use. This technology can make sense of all of that data in rapid time and with greater accuracy.
When man invented the first computer back in the 1940s (the ENIAC or Electronic Numerical Integrator and Computer) the goal has always been to create a machine that was capable of human learning and thinking.
Unfortunately, we don’t have a machine that thinks like a human just yet. But we have something that is close to it and is beneficial to us.
Types of Machine Learning
There are three types of machine learning, which are:
- Supervised Learning (task-driven predicting the next value)
- Unsupervised Learning (data-driven and works by identifying clusters)
- Reinforcement Learning (the ability to learn from mistakes)
In this type of ML, the data being interpreted and processed is systematically labeled. This labeled data is used to train the algorithms used by software. Under this type of ML are two types called regression and classification.
In this type of ML only unlabeled data is used to train an algorithm. The goal here is to train the algorithm to find the structures within the data and explore it.
The algorithms used in supervised learning are designed to classify data into various segments. It will then create new data segments as it organizes data that it encounters and it will then create new labels.
In this type of ML no raw data is used. The goal here is to make the algorithms of a program to figure situations out on their own. This is the ML that is used for gaming, navigation, and robotics. Programs and systems created with reinforcement learning tend to learn from their mistakes.