ప్రేమ వెన్నెల| telugu lyrics Prema vennela song lyrics in telugu|

Introduction To Artificial Intelligence.

Hello guys, how are you doing..?

Let’s start with AI for programmers. I pinned hello world program in story title because this post is for beginners so we are not going into much details so beginners can easily understand the core idea behind AI Algorithms and how it works from a programming perspective.Before we start coding first let’s clear some theory about machine learning, this era is all about AI(artificial intelligence). we heard many things like deep learning, neural networks and what not. For beginners these terminologies create confusion. So, let me first clear full picture and some of these terminologies in simple terms.

FIRST STEP towards AI.

AI. is one of the departments of CS(computer science). AI is all about taking much complex and harder problems that traditional programming will not be able to solve, these problems are solved by simulate human intelligence. For example, neural networks are a replica of the human brain. Hey, wait I don’t know about neural networks, I am not a biology student… hang on guys we will dig step by step.

Then what is machine learning and deep learning, how it differs from AI…?

Machine learning is a subset of AI. Trust me guys there is more in AI. , but we can say that machine learning is a most popular stream of AI. Machine learning contains many popular algorithms for solving complex problems.

A question should arise why machine learning, AI and why I am writing this post, this stuff we have already from the very beginning of CS. Most of the algorithms are from 19’s then why now..? because now today we have more data for analysis, predict and more computing power that we never had before.

Note: All image borrowed from www

Machine learning algorithms or maybe we can say problems that we want to solve using machine learning that is classified into mainly three terminologies Supervised learning, Unsupervised learning and Reinforcement learning.

Supervised learning — In machine learning practice most of them are supervised learning, in supervised learning, we have input feature-set (variable) X and output label (class) Y. We just want to find mapping between X and Y with an algorithm. So, Y = f(X) in very simple term if we have an image and its true labels for training. Then it will be supervised learning problem.

Further, supervised learning can be grouped into classification and regression,

Classification-: Classification problems is when our output Y is always in categories like positive vs negative in terms of sentiment analysis, dog vs cat in terms of image classification and disease vs no disease in terms of medical diagnosis.

Regression -: Regression problem is when we want to predict some real value from existing sequence of values. Like we have series of points(numbers) that represent stock market price of any particular company and we want to predict next points (numbers) from given points. Thus stockmarket prediction is regression problem.

Unsupervised learning  — As exact opposite of supervised learning, we have only feature set(variables)X that we have to classify only using their features so here, we don’t have true labels of their training samples. For example, we want to classify dog vs cat but we don’t have their true labels we just stack all pictures of dogs vs cats and find two clusters or groups that classify them with their feature values(pixel intensity).

Supervised learning needs a huge size of handmade labeled data-set and we don’t have labeled data for all problems so most of the real time problem can be solved with unsupervised learning.

Unsupervised learning can be grouped into clustering and dimensionality reduction.

Clustering:- In clustering, we grouped all the data into different clusters upon there behaviour and features.

Dimensionality reduction:- If your number of features is high in samples, then it may be useful to reduce it with an unsupervised algorithm that can be used to reduce the dimensionality illustrate in the following image. So further it can be easy to classify also dimensionality reduction is useful to visualize data because highly dimensionality data is hard to plot.

Reinforcement learning  — Reinforcement learning is the hottest topic in AI. It is the way to accomplish true AI. where algorithms find path towards the goal using trial and error, no prior knowledge about a problem is required. An agent just has to monitor the environment and find a path towards goal. For example, here we assume game environment as the simulation of real world and game’s character has to jump for passing different obstacles and reach to the goal state. So that character from the game is our agent in reinforcement learning. And we give score +=1 for each successfully passed obstacles and agents job is to gain more scores and reach the goal. Let’s take another example to understand the clear picture, assume we want to train a dog for doing any task, we cannot teach dog directly so we are given a reward to a dog for every right moves and punishment for wrong moves. This way dog has to figure out what it did that made it get the reward or punishment. And next time it will know that what step gives him rewards and what gives punishment.In a similar way, we can train to computers also and accomplished many problems like controlling robots, train game bot and so on.

Deep learning:- We can say that deep learning is one kind of machine learning that gain more and more attention nowadays because of its success in different fields like, computer vision to signal processing and medical diagnosis to self-driving cars. As all other AI algorithms deep learning is from decades, but now today we have more and more data and cheap computing power. So how deep learning differs from machine learning???

Let’s take our dog vs cat example if we want to solve dog vs cat classification problem with machine learning algorithms we have to do handmade feature extraction, like define how cats look like and how it differs from the dog. if we encounter this same problem with deep learning, deep learning algorithms like CNN(Convolution neural network) does feature extraction automatically. The same process it works for signal processing and also other problems. What is CNN, wait for next post guys. we will not cover everything in single post…!!

So here, if our universe is A.I. Then our Solar system is machine learning and all planets are different machine learning algorithms and structures. Further, special planet Earth will be Deep Learning because it gains more success and fame than others.