Have you ever thought how Facebook recognizes people’s faces in the picture you uploaded? Or sometimes you might have noticed it suggest you people on the basis of your friend list or on the basis of profiles you visited. These all are possible due to machine learning. There are many more things that is done by the help of machine learning. But what is machine learning.
Machine Learning is the part of artificial intelligence which makes data driven decision, can learn and formulate predictions based on some past experience by applying algorithmic and analytic model to preprocessed data. It is basically an idea where computer program can learn and adapt to new data with no human interference which means, we should be able to give machine the access to data and let them learn from themselves. These programs are designed to learn and improve over time when exposed to a new data.
Why do we need machine learning?
Machine learning has changed our life in several significant ways. The reason why we need machine learning are as follows:
1. Powerful Processing: It allows powerful processing which means, we can process through far more complicated data. Also the processing is quicker. That means, we can do more work in lesser time.
2. Better decision making and prediction: The decision made will be much better and the predictions are much more accurate.
3. Accuracy: The obtained output is more accurate by the help of machine learning.
4. Analyzing complex data: When we get more and more data it gets way too complex. Through machine learning we can analyze complex big data.
Life before Machine Learning:
1. Before machine learning we had rule based AI system. These systems failed in real life scenario. As the programs failed at the test scenarios. Moreover, they were not so useful in real life application.
2. If we look back 5–6 years, facial recognition was only shown in sci-fi movies. We had to take the trouble of tagging every single person present in our picture.
3. There were no voice recognition system like Siri and Cortana.
Types of machine learning:
As said earlier, machine learning allows us to built self learning machines that evolve by itself without human aid. Based on user behavior, data patterns and past experience it makes important futuristic decisions. There are three types of machine learning:
1. Supervised Learning: Supervised learning is the method where the model (after sufficient training) is able to make prediction. This algorithm consists of result which is to be predicted from a known set of predictors. The best way to understand supervised learning is considering the example of Gmail. When we receive any mail and we know that the mail is spam, then we simply mark it as spam. Now the Gmail will automatically put similar mails in the spam box by spam filtering method.
2. Unsupervised Learning: It deals only with the input data by ensuring that the data is organized and readable. It analyzes the input data for finding out patterns. Observation is the basis of its learning. In this, we don’t have any target to predict. It is basically for finding hidden pattern.
3. Reinforcement Learning: In this the machine learns from past experiences and tries to capture the best potential knowledge for making accurate business decisions. Reinforcement learning has three stages. For this we will take the example of chess game. In the game of chess the machine makes a move, this is the decision (first stage). After that it gets to know whether the move was good or bad, this is the feedback (second stage). At last, it learns from the feedback (third stage). The mobile and computer games are made on the basis of reinforcement learning. It involves iterative learning based on the past results.
Steps involved in machine learning:
1. The first thing we need to do is to define an objective. Once we defined an objective then we proceed with collecting data as we can’t train our machine without data. We need to collect certain data. The data can be in the form of text, audio, video or images.
2. We clean the collected data by ensuring that data has all useful field. We clean the data because sometimes, the data that is collected is not in the form to be processed further. Data can have some missing values or some errors. So, it has to be rectified.
3. We then select appropriate algorithm.
4. We train the model with the algorithm based on the analysis.
5. After this we do the testing by providing certain set of data. The actual output is known prior to test. The test is made to ensure whether the output generated by the algorithm is equal to the actual output or not.
6. After this our trained model and new data are put together for making prediction.
Real life Applications:
There are many real life applications of machine learning. Let us look on some of them:
1. Google Maps: It analyzes speed of traffic through anonymous location from our smartphone. This enables Google to suggest best possible route in context of time and traffic.
2. Image Recognition on Facebook: When we upload a picture on facebook, the facebook is able to recognize the people in the picture. It takes the pictures of people as data set and then compasses the features and faces of people present in our picture. Hence identifying who’s in our picture. This is an example of unsupervised learning.
3. Netflix: Suppose we watched game of thrones. Now the Netflix will suggest us Lord of Rings movie. Netflix uses recommendation engine to recommend us series based on the series watched by the user.
4. Video Surveillance: How boring and difficult it would be for a person to sit and monitor several cameras. The video surveillance helps in detecting crime before they can happen by tracking the unusual behavior of people such as, standing motionless for long time, stumbling etc.
5. Email Spam: The system security is nowadays powered by machine learning that can understand the coding pattern. Therefore they detect spam mails.
6. Google search engine: Google and other search engines uses machine learning to improve the search results for us. Every time we search the algorithms keep a watch on our reaction to the results.
7. Fraud detection: Machine learning provides us secure cyberspace by tracking monetary frauds. It gathers and segments the data. Then the ML model is fed with training sets for predicting the probability of fraud.
8. Biometric Identification: In this we train the machine and after a couple of biometric identity the machine can validate our future input and will identify us.
So, this was the basic introduction to machine learning.