Artificial Intelligence, Machine Learning and Deep Learning
Technology has been upgrading itself at a scale undreamt off. Every day, every hour, every minute we have something new. Data Science is one of the few new buzzwords of this technological phase. Research states that there has been an immense influx of data post-2008. Every second, numerous system all over the world process terabytes of data. This has resulted in the development of the field of data science.
Artificial Intelligence needs no introduction. Everything around us is indirectly or directly an implementation of an Artificial Agent. An example? YouTube.
YouTube has played a significant role in our lives. Where do you think the suggested content comes from? YouTube checks out each user’s preference of videos, figures out a pattern and hence throws forth videos that the user may be interested in. This is not the work of any person sitting and checking your preferences. Need a break?
Ask Siri, she’s probably got something :)
Yes, one of the most widely used agents in the world is Apple’s Siri. Siri is a virtual agent, with no intelligence of its own. Yet, she can get you anything you want. How does this work?
A group of data scientists got together and decided to implement what we now term as artificial agents. These guys are pretty good when it comes to technology; besides that, they need to feed the agent data. They need it to understand how decisions and predictions are supposed to be made. Hence, a very strong mathematical expertise is crucial for a successful agent. Mathematical expertise refers to algorithms that give predictive results. This process refers to Machine Learning. A machine is being fed these effective algorithms and is made to understand that this is needed to solve problems.
Confused? Yup, relatable. Often, many confuse AI with ML. Machine Learning Algorithms can be related to the intelligence that is within us humans to be able to make decisions and arrive at conclusions. Artificial agents are those that have this “intelligence” fed to them so that they can enact human behaviour.
Let’s dive a bit into the technicalities of Machine Learning. More technically, Machine learning involves algorithms that enable computers to learn from a set of data and improve themselves with almost 0 human intervention.
It is broadly classified into 3 types:
1) Supervised Learning.
2) Unsupervised Learning.
3) Reinforcement Learning.
1) Supervised Learning Algorithms:
In a supervised learning model, the system is given some data which is labelled by the user. What that means is that each data value included has the correct label. The final aim here is to make sure the mapping function involved can predict values for certain data with a minimal amount of variation.
TYPES OF SUPERVISED LEARNING:
· CLASSIFICATION: A classification problem deals with true or false values. Example: If you’re checking for disease, classification can be done as positive or negative for the disease.
· REGRESSION: A regression problem is when a certain relation has to be mapped between the output and input variables, and results need to be computed.
2) Unsupervised Learning Algorithms:
In unsupervised learning, the system/algorithm is given unlabelled and uncategorized data. The algorithms are designed to act on the dataset fed without any prior training or teaching.
To give a very basic example, let’s take our favourite cartoon characters.
In this example, the model is fed the following characters. The model makes an assumption based on whether the following character is a duck or not. There is no label or category provided in the training data. The model can make this separation by looking at the data and using various algorithms to make the difference.
TYPES OF UNSUPERVISED LEARNING:
· CLUSTERING: A clustering problem is where one wishes to group-specific data showing similar properties. One typical example can be grouping people based on their favourite ice cream.
· ASSOCIATION: An association rule learning algorithm is where you want to find if there’s a similarity in the data. Example: People who like Vanilla Ice cream also like to have chocolate Ice cream.
Reinforcement Learning
3) Reinforcement Learning Algorithms:
In reinforcement learning, the algorithm or smart agent learns by interacting with the environment it’s subjected to. As a reward, the agent is given rewards for correctly performing the task and penalties are deducted accordingly for an incorrect result. This is a type of dynamic programming that makes sure algorithms are trained using this reward and punishment system. In the above example, we see that our dear Dino needs to navigate the maze. For every correct turn, it’ll be awarded points and points would be accordingly deducted for moving in the wrong direction.
Reinforcement learning is implemented through Deep learning networks which will be covered later on.
That’s it for now folks !!
Hope you liked it.
Team FACE.
In C<>de, we trust.
