Comic Edition: What is Artificial Intelligence? What is Machine learning? What is Deep Learning? Comic Edition: What is a Neural network?

This machine learning basics video will help you understand what machine learning is, what types of machine learning there are — supervised, unsupervised, and reinforcement, how to learn machine learning through simple examples, and how to use machine learning in various industries.

What is machine learning?

We know that humans learn from past experiences and machines follow human instructions

How do humans learn? By inputting certain information to the brain, knowledge and experience can be learned and summarized, and decisions or actions can be made based on existing experience when similar tasks are performed.


Simple example

Paul listens to new songs and decides whether he likes or dislikes them based on the rhythm, intensity and gender of the voice.

Yeah, Paul loves this song.

By looking back at Paul’s past choices, it’s easy to categorize unknown songs. Suppose now Paul hears a new song, let’s label it B, and B is somewhere here, medium tempo, medium intensity, neither relaxed nor fast, neither soft nor flying.

Now can you guess whether Paul likes it or not? It is impossible to guess whether Paul will like it or not, and the other options are unclear. Yes, it’s easy to categorize song A, but when the choice gets complicated, it’s like song B. Machine learning can help you solve this problem.

Let’s see what happens. In the same example of song B, if we draw a circle around song B, we see that there are four green dots for liking and one red dot for not liking.

If we choose the majority of green dots, we can say that Paul will definitely like this song, and this is a basic machine learning algorithm, it’s called the K-nearest Neighbor algorithm, and this is just a small example of one of the many machine learning algorithms.

But what happens when the choices get complicated? Like in the case of song B, when machine learning comes in, it learns the data, builds predictive models, and when new data points come in, it can easily predict it. The more data, the better and more accurate the model.

Classification of machine learning

There are many ways of machine learning. It can be supervised, unsupervised or reinforcement learning.

Supervised learning

Let’s start with a quick look at supervised learning. Suppose your friend gives you a million coins of three different currencies, say one euro and one euro, each with different weights. For example, a one-rupee coin weighs three grams, a euro weighs seven grams, and a euro weighs four grams. Your model will predict the currency of the coin. Here, weight becomes the feature of the coin and currency becomes the tag, and when you feed that data into a machine learning model, it learns which feature is associated with which outcome.

For example, it will learn that if a coin is three grams, it will be a rupee coin. Based on the weight of the new coin, your model will predict the currency. Therefore, supervised learning uses label data to train models. Here, the machine knows the characteristics of the object and the labels associated with those characteristics.

Unsupervised learning

At this point, let’s look at the difference with unsupervised learning. Suppose you have cricket data sets for different players. When you send this data set to the machine, the machine recognizes patterns in player performance, so it processes the data with its respective Achatz on the X-axis, while running on the Y-axis

When you look at the data, you can clearly see that there are two clusters, one cluster is players who score high and score low, and the other cluster is players who score low and score high, so here we interpret the two clusters as batters and bowlers.

It is important to note that there are no labels for batters and pitchers, so learning using unlabeled data is unsupervised learning. Thus, we learn about supervised learning with labeled data and unsupervised learning with unlabeled data.

Reinforcement learning

Then there’s reinforcement learning, which is reward-based learning, or we could say it works by feedback.


Here, suppose you feed the system an image of a dog and ask it to identify it. The system recognizes it as a cat, so you give the machine negative feedback that it’s the image of a dog, and the machine learns from that feedback. Finally, if it comes across any images of other dogs, it will be able to categorize correctly, and that’s reinforcement learning.

Let’s look at a flow chart, input to the machine learning model, and output based on the applied algorithm. If it is correct, we take the output as the final result, otherwise we provide feedback to the train model and ask it to predict until it learns

Application of machine learning

You sometimes don’t know how machine learning is possible in this day and age, and that’s because today we have a lot of data available, everyone is online, either trading or going online, generating a lot of data every minute, and data is key to analysis.

In addition, the memory processing power of computers has been greatly increased, which helps them process such large amounts of data without delay.

Yes, computers now have a lot of computing power, so there are lots of applications for machine learning.

Machine learning is used in health care, where doctors can predict diagnosis, mood analysis, just to name a few.

Social media recommendations by tech giants are another interesting application. Machine learning fraud detection in financial sector and forecasting customer churn in e-commerce sector.

quiz

I hope you have an understanding of supervised and unsupervised learning, so let’s do a quick quiz to determine whether supervised or unsupervised learning is used for a given scenario.

  • Scenario 1: Facebook identifies your friends from a tagged photo album
  • Scene 2: Netflix recommends new movies based on someone’s past movie selections
  • Scenario 3: Analyze bank data for suspicious transactions and flag fraudulent transactions

Scenario 1: Facebook identifies your friends in a photo album of tagged photos Explanation: This is supervised learning. Here, Facebook is using tagged photos to identify the person. So the tagged photo becomes the tag of the picture, and we know that when the machine learns from the tagged data, it is supervised learning.

Scene 2: Recommending new songs based on someone’s past music choices Explanation: This is supervised learning. The model trains classifiers on pre-existing labels (song genres). This is what Netflix, Pandora, and Spotify have been doing, collecting songs/movies you already like, evaluating features based on your preferences, and then recommending new movies/songs based on similar features.

Scenario 3: Analyzing bank data for suspicious transactions and flagging fraudulent transactions Explanation: This is unsupervised learning. In this case, suspicious transactions are not defined, so there are no “fraudulent” and “non-fraudulent” labels. The model tries to identify outliers by looking at outlier transactions and flag them as “fraud.”