Machine Learning.

Machine Learning is the next Internet — Tony Tether

We uses Machine Learning (ML) numerous times in a day without even knowing it. Like every time we hit Netflix we uses its recommendation system, when opens our mobile it recognizes using face/finger prints. Each time we open our mail we see some in our inbox and few in our spam folder that because of machine learning. All these and many more are the examples of machine learning we uses in our day-to-day life. So in this article we’ll get to know what is ML, why it is used and how does it works.

What is Machine Learning?

The term “Machine Learning” is coined by — Arthur Samuel.

Machine learning is the branch of computer science and subset of artificial intelligence (AI). It is focused on use of data and algorithms to mimic like humans and improve with experience without being manually programmed. Using statistical methods, algorithms are trained to make predictions by finding patterns within data.

Importance of Machine Learning

The availability of limitless quantity of data, affordable storage increased the growth of machine learning lately. Many industries are developing more robust machine learning models that can handle complex data very easily and in faster way with more accuracy. Machine learning tools help industries identify profitable opportunities with no time, and reduce risk.

Importance of Data

Machine learning starts with data — data can be anything like numbers, texts, photos, time series data, speech etc. The data is gathered, cleaned and prepared for feeding to the machine learning model for the training. The more the data is the better the outcome/prediction.

The benefits of machine learning

Humans can typically create one or two good models a week; machine learning can create thousands of models a week. — Thomas H. Davenport

Machine learning is constantly increasing and when someone/something increases, it increases because of its capabilities. And so is ML.

• ML helps identifying patterns within data (structured and unstructured).
• ML is excellent at data mining and improve with time.
• ML detects fraud by monitoring and identifying new patterns and catch attempts before they’re successful.
• ML can save huge amount of time, and man power.

How does machine learning works

Depending upon the type of the data and the required outcome, particular machine learning models can be used. Machine learning algorithms are basically used to classify things, predict patterns, and make decisions. Algorithms can be used one at a time or combined (ensemble technique) to achieve best out of it.

Machine Learning Methods

There are four types of machine learning methods which we use.

1. Supervised Learning

In Supervised learning, the algorithm is trained by the labeled data to classify data or to predict outcome. Supervised learning models consist of ‘input’ and ‘output’ pair, where the output is labeled with desired value. In simple words, in supervised learning the algorithm is trained under supervision i.e., we give each value some labels and our model gets trained with given data and provides us required outcome.

Supervised learning is used in recommendation systems that we uses is our day-to-day life, etc.

Some methods used in supervised learning includes -

• Linear Regression
• Logistic Regression
• Random Forest
• Support Vector Machine
• Neural Networks and more.

2. Unsupervised Learning

In Unsupervised learning there is no labeled data. The machine learns the behavior of input data (unlabeled and unstructured) and tries to identify correlation, using the accessible data. Unsupervised learning uses algorithm to analyze and cluster unlabeled data. As we humans uses our intuitions and experience to acknowledge something likewise unsupervised learning do the same to find pattern within the data to predict accurate outcome.

Unsupervised learning is used in facial recognition, cybersecurity, etc.

Some methods used in unsupervised learning includes -

• Neural Networks
• k-means clustering
• Hierarchical Clustering
• Principal Component Analysis and more.

3. Semi-supervised Learning

Semi-supervised learning lies between supervised and unsupervised learning. Semi-supervised learning uses a smaller labeled dataset for classification and feature extraction from unlabeled dataset. The labeled data is used to initiate the process and can considerably improve learning speed and accuracy. The semi-supervised learning can solve the problem of not having enough labeled data for the prediction. The semi-supervised learning algorithm instructs the machine to analyze the data that could be applied to unlabeled dataset.

Semi-supervised learning is used in linguistic and speech analysis, fraud detection, etc.

4. Reinforcement Learning

Reinforcement learning is similar to supervised learning, the only difference is supervised learning uses labeled data to predict output while in reinforcement learning we uses hit and trial method. In simple words, model learns by making mistakes in reinforcement learning.

For example, we cannot label every move while playing chess. We learn chess by using trial and error techniques, every time we make mistakes we gets penalty (by losing one of our troop) likewise for every right move we get rewards (by defeating opponents one troop).

Reinforcement learning can be used in computer game development, stock market trading, etc.

Conclusion

I hope you learned something from this post, if I forgot to add something feel free to add your thought. Feedback will be appreciated, so don’t forget to leave comment about your first thought after reading this article. Thanks!

Before leaving here’s a very interesting quote:

A baby learns to crawl, walk and then run. We are in the crawling stage when it comes to applying Machine Learning — Dave Waters