February 16, 2017
The goal of Artificial Intelligence is to create a machine that can mimic a human mind and to do that, of course, it needs learning capabilities, however, it is more than just about learning, it’s about reasoning, knowledge representation and even things like abstract thinking. Machine learning, on the other hand, is solely focused on writing software that can learn from past experiences. One thing you might find astounding is that Machine learning is closely related to data mining and statistics than is to AI. Why is that, first we need to know what we mean by Machine learning.
Machine learning is a method of data analysis that automates analytical model building. Using algorithms that iteratively learn from data, machine learning allows computers to find hidden insights without being explicitly programmed where to look.
If a computer program can improve how it performs a certain task based on past experience than you can say it has learned. This is quite different to a computer program that can perform a task because it is programmer has already defined parameters and data needed to perform that task. For example, a computer program can play Tic-tac-toe because a programmer has programmed it with a built in winning strategy. However, a programmer does not have a pre-defined strategy and only a set of rules about the legal moves we need to learn are repeatedly playing the game until it is able to win. This doesn’t only apply to games, it’s also applicable to programs which perform classification of prediction, classification is a process by which a machine can recognize and categorize things from a data set, including visual data and measurement data. Prediction, known as the regression in statistics is where a machine can guess or predict the value of something based on some previously given values. For example, given a set of characteristics about a house, how much is it worth, based on previous house sales. This leads us to another definition of machine learning. It is the extraction of knowledge from data. You have a question you are trying to answer and you think the answer is in the data that is why Machine learning is related to statistical analysis and data mining.
Machine learning can be split into three categories Supervised learning, unsupervised learning and reinforcement learning.
Supervised learning is where you teach train the machine using data which is well labeled. It means that the data is already tagged with the correct outcome. Here is a picture of the letter ‘A’, this is a flag for India, it has three colors, and one of them is white and so on. The greater the data set the more the machine can learn about the subject matter. After the machine is trained, it’s given new previously unseen data and a learning algorithm then uses the past experience to give you an outcome. This is the letter ‘A’. That is the Indian flag and so on.
Unsupervised learning is where the machine is trained using a dataset that doesn’t have any labels. The learning algorithm is never told what the data represents. Here is a letter, but no information about which letter it is. Here is a characteristic of a particular flag without naming that flag. Unsupervised learning is like listening to a podcast in a foreign language which you do not understand. You do not have a dictionary and you don’t have a supervisor to tell you what you are listening to. If you listen to just one podcast it won’t be much benefit to you but if you listen to hundreds of hours of those podcasts your brain will start to form a model of how a language works. You will start to recognize patterns and you will start to expect certain sounds. You will learn that language much quicker when you do get hold of a dictionary or a tutor.
Reinforcement learning is similar to unsupervised learning in that the training data is unlabeled. However when asked a question about the data the outcome will be graded. A good example of this is playing games. If the machine wins a game than the result is backed trickled back down through the set of moves to reinforce the validity of those moves. This isn’t much use if the computer plays just one or two games but if it plays thousands or millions of games than the cumulative effect of the reinforcement will create a winning strategy.
There are many different techniques for building Machine learning systems and many of these techniques are related to data mining and statistics.
One of the buzzwords we hear from companies like Google and Facebook is neural net. A neural net is a technique of machine learning modeled on the way neurons work in the human brain. The idea is given a number of inputs the neuron will propagate a signal depending on how it interprets those inputs, in Machine learning terms this is done by matrix modification along with an activation function. The use of neural networks has increased significantly in recent years and the current trend is to use a deep neural network with several layers of interconnected neurons. During Google I/O 2015 it was explained how much machine learning and deep neural networks are helping Google to fulfill its core mission to organize the world’s information and make it universally accessible and useful. To that end, you can ask Google now things like how do you say “How are you?” in Spanish? And because of neural networks, Google is able to do voice recognition, natural language processing, and translation. Currently, Google is using 30 layered neural nets. As a result of using these neural networks, Google’s error rate for speech recognition has dropped by more than 30%.
Machine learning enables cognitive systems to learn reason and engage with us in a more personalized and natural way. These systems get more customized and smarter through interactions with data, devices, and people. With all the information that surrounds us, they will help us take on what may have been seen as unsolvable problems and bringing the right insight or suggestion to our fingertips right when it is most needed.