Machine Learning (ML), categorized under Artificial Intelligence, allows machines to learn with experience or pattern.
Makes sense? Remember “The Terminator”, the movie? Let’s just say that the movie demonstrates ML and Artificial Intelligence perfectly. So, With such as #Machinelearning, we can make a #machine program other machines. Scary, right?
Artificial Intelligence such as Machine Learning is similar to human learning – a journey from childhood to adulthood through a series of mistakes and experiences. Machine Learning has created quite a buzz and is certainly not overhyped. In fact, almost all the major players in IT (Facebook, Apple, Microsoft, and Google) are using ML in their product in one way or another.
Through the following post, let’s build some intuition about Machine Learning and its application.
To get you excited, let’s look at some real life applications of ML.
Robotics: As was witnessed in “The Terminator”, the machine had become a threat to the human race. How did the machine learn? The best plausible answer is through a neural network schema built using advanced Machine Learning algorithms. The application of ML is enormous in the field of robotics. Some other instances are robots from Boston Dynamics and Humanoid robots.
Image Processing: Image processing applications include face detection on Facebook, image comparison in Google, Instagram filters, and virtual reality.
Speech and Text Recognition: Auto correct, Chat bots, and Sentiment Analysis.
Data Analysis: Grouping, prediction, and association rule.
How does it work?
Imagine you need to create an algorithm for sentiment analysis on twitter. We know that there are certain words which can be categorized based on sentiment, for instance Positive = {“Win”, “Awesome”,” Enjoy”} and Negative = {“Loss”, “Bad”, “Sad”}. A search for these words in the twitter posts would help provide a score based on the number of words matched and the degree of positivity and negativity.
Based on the score, the machine can figure out the sentiment of a Twitter post.
Machine Learning Algorithms Category
Supervised Learning: It is an algorithm which uses labeled data as a training set. Based on a hypothesis, it will generate output as a prediction. Some other forms of supervised learning are Classification and Regression.
Unsupervised Learning: This algorithm does not use any predefined hypothesis for input data. Additionally, there is no label to categorize the data. The algorithm needs to find out the hidden pattern and its structure to make sense out of it. Some instances of Machine Learning algorithms under Unsupervised Learning are clustering and K-means, among others.
Machine Learning Algorithms
- Probabilistic Models: This algorithm uses probability distribution to predict output, e.g. Naive Bayes or Bayesian Methods.
- Deep Learning: It is neural network based algorithm which use Artificial Intelligence to implement. It is similar to a human who builds an understanding by referencing and creating layers of information based on experience to analyze patterns or logic.
The above Machine Learning algorithms are the two most commonly used algorithms out of the numerous Machine Learning algorithms available.
How to choose an algorithm?
- Opt for supervised learning if you’re dealing with labeled data which already has results associated with it, enabling output prediction based on the predefined pattern. Otherwise, go for unsupervised.
- If you are looking to create a recommendation system like the one in Amazon, then Clustering is an ideal choice.
- If you are dealing with a discrete set of data, ensure classification.
- If you’re dealing with a score on prediction, then regression is an essential element.
I hope that you were able to glean valuable insights and understanding about Machine Learning. Keep checking this space for more extensive description and code examples.
