Machine Learning (ML) and Deep Learning (DL) are the two of the hottest buzzwords after Artificial Intelligence (AI) at the moment.
Our computers have evolved tremendously in the last couple of decades. By evolution, I don’t simply mean the advancements in processing power and data storage capacity, but also changes in the way machines function overall.
This has allowed computers to learn and perform increasingly complex tasks. A famous example of this — IBM’s Deep Blue computer successfully defeated reigning world Chess Champion Garry Kasparov in 1997!
But what is it that makes a computer efficient yet smart enough to perform such unbelievable and complex tasks?
Let’s find out!
What is Machine Learning?
All email providers today offer a special mail category known as ‘Spam’, but how does the email program figure out the criteria for this type of categorization?
The answer lies in Machine Learning.
The magic happens with a set of algorithms that are devised and implemented using a database of common phishing/irrelevant/misguiding/unwanted terms &/or phrases.
It could also be based on certain specific/typical patterns of email releases or rollouts. So when you receive a spam email, machine learning algorithms at the backend enable the system to identify and sort such messages automatically, placing it in the Spam folder and saving you the annoyance of doing it yourself while keeping you safe online.
I often receive emails that claim I have won the lottery. As tantalizing as that may be, I know for a fact such emails are bogus. Partly, because I avoid lotteries, and mostly because these emails contain a link to a malicious page.
Thankfully, my email provider is usually able to chauffeur such mails into the spam folder, keeping me and my computer safe. But, how does it do that?
Machine Learning (ML) algorithms scan such emails for words like “You have won the Lottery! Please click here to claim your prize.” They also verify whether the same mail has been sent to several users (thereby detecting a pattern). If these conditions are met, the rogue mail is relegated to the spam folder.
The awesome bit about ML isn’t that it can do this. It can also continuously evolve over time becoming more robust and data-driven.
Over time, it continuously parses data and learns from patterns thereby enabling better probabilistic prediction(s) correcting previous erroneous patterns in its repository. In English, this means it learns and gets better over time! Yay!
Typically, this requires the algorithm to learn from a huge amount of data (called a dataset) using it to analyse and improve predictions.
But, I cannot speak about ML without Artificial Intelligence (AI) which is a key factor driving ML. Why? ML algorithms are mostly derived using the fundamentals of AI.
There are several ML algorithms. Let me list out the most widely used ones:
1. Ensemble — Random Forest, Bagging and Boosting
2. Neural Networks — Perceptron, Back-Propagation
3. Regularization — LASSO and Ridge
4. Regression — Linear Regression, Logistic Regression, Least square methods
5. Clustering algorithms — k-Means Clustering, Expectation Maximization
6. Instance Based — k-Nearest Neighbour
7. Dimension Reduction — Principal Component Analysis, Linear Discriminant Analysis
8. Decision Tree — Classification and Regression Tree
9. Bayesian — Naive Bayes
What is Deep Learning?
The next step in the evolution of Machine Learning is Deep Learning or DL. DL is an extended part of the ML Family that enables more efficient ways to implement ML. However, I cannot speak about DL without talking about Artificial Neural Networks or ANN.
Chronology of evolution of Machine Learning
Machine Learning → Artificial Neural Networks → Deep Learning
An Artificial Neural Network is defined as :
“Artificial Neural Networks, are computational models based on the structure and functions of biological neural networks. It is like an artificial human nervous system for receiving, processing, and transmitting information in terms of Computer Science.”
An ANN consists of layered systems in which each layer has several building blocks known as Neurons which perform specific computation task(s). The idea is to mimic the biological nervous system.
Broadly, these layers are divided into 3 — Input Layer, Hidden Layer and Output Layer.
Deep Learning automates the task of predictions i.e. it helps design a model through which we can pass our dataset.
Why is this useful?
It is useful because in return it automatically computes all features/patterns significant for analysis and predictions. I think of it as a digital Sherlock Holmes, able to look at and identify patterns to discern a result.
The 5 most commonly used DL models:
1. Generative Adversarial Networks
2. Convolutional Neural Networks
3. Recurrent Neural Networks
4. Auto Encoders
5. Recurrent Neural Network
The DL methods deep learning engineers must be aware of:
1. Back Propagation
2. Learning Rate Decay curve
5. Stochastic Gradient Descent
6. Batch Normalization
7. Long Short-Term Memory
Machine Learning came into existence to develop smart systems and Deep Learning (along with ANN) took it to a different level altogether. While Machine Learning was just the beginning, Deep Learning has enhanced it to the point where we have humanoids and self-driven cars which were fictional just a few years ago!
I hope this blog helped you get some clarity on deep learning and machine learning. Always remain hungry for deep insights and be open to research!
Thank you for reading!