Machine Learning [In Nutshell]

Machine Learning is the word we all see in movies recently, they show some cool stuffs robots fighting, cloning human beings but when we dig deep all there is nothing but data’s and algorithms those where not cool as we see in Avengers and IronMan. Sorry to break your spirit but the beauty is mathematics and formulas that is the magic spell off Machine Learning.

Whats is Machine Learning

Let the machine to classify the data and make decisions or analysis the data that human cant process it the rate we cant imagine. In nutshell let machine think. So how a dumb machine that knows 0 & 1 to think like a human? secret sauce is Algorithms aka, Mathematical formulas so you may guessed mathematics gives yes/no or right/wrong, so that’s the way machine take decisions in a rate that humans cant do it.

Mathematics: The Beautiful Language of the Universe

Classification of Machine Learning

We have problems like with big data, so how can organize the problems and solutions. We classify into two as below,

  • Supervised Learning
  • Unsupervised Learning

Supervised Learning

We deal with the problem that we know the final result or at least the probability of the result but we cant process a huge amount of data that will take ages to process by humans. Scenarios we deal with as below,

  • Formatted Data
  • Probability of result is already known
  • Analysis of data
  • Solve known problems
  • Example:Stock market

Unsupervised Learning

We have pica bytes of data and its from different sources that we cant humanly possible to organize it and process it those problems deals as Unsupervised Learning.

  • Let machine organize the data
  • Data is complicated so its hard to find results
  • Solve unknown problems
  • Discovering problems and solutions
  • Example:Astronomy
  • Soon weill write about data processing, Stay tuned.

AI – Uncertainty Management

Uncertainty in AI to make decision in situation like,

What to do? when we don't know what to do.

Its kind of tong twister though but its a situation we all came across as a developer, we need to handle the exceptions in the code[but not all developers] in which we don’t know handle some exceptions when it arise but we will just place it in try… catch block as we too lazy to find the root cause. But in AI we need to take the decision to keep the process up and running in situation which occurred by factors like,

  1. Sensor limitation – limitation when we try to use one sensor for two operation.
  2. Adversaries – Make hard to understand the environment.
  3. Stochastic environment – Random outcome of events.
  4. Laziness – lazy to computing sensor data.
  5. Ignorance – We fail to handle the situation that may or may not occur.

Machine Translations:

As the translating engine is trained with different examples of two different language with rater not difference between stokes and letters of two language, it will learn about the related words for each words in the languages so it can match the words on translating process. It’s more like photographically memory in the engine.

Lets break down the process in programming language,

  • Array ‘A’ which has words in Hindi.

  • Array ‘B’ which has related Hindi words in English.

  • Now all we need to do is match the following words, a simple switch case can do the job.

As we keep adding the words the code need to be refined and handle those situation, this is where AI plays its part by learning the process and make it better by probability of chances the word related to other language.

Terminology in AI

Terminology in AI which deals with different scenarios of problem that an Intelligent Agent AKA AI need to handle as follows,

  1. Fully VS Partially Observable Environment:
    • Fully Observable Environment:
      > An agent can get complete sufficient data about the environment any given time to take optimal decision.

    Example : In card game in which agent can predict the card in persons hand at any given time.

    • Partially Observable Environment:
      >An agent can get only partial data about the environment and store the past and present state of the environment in a memory to take optimal decisions, This is the mostly often scenario in AI we need to handle.

    Example : In poker agent can predict the game as the persons past and present state to predict future move.

  2. Deterministic VS Stochastic Environment:

    • Deterministic Environment:
      > Environment that can be predicted the next state in any given time as the event are predictable

    Example: Chess, where ever the position of the piece still we can determine the next move of the opponent.

    • Stochastic Environment:
      >Environment that we cant predicted the next state as the outcome where random in nature.

    Example: Dice, As the game has random nature of dice that we cant predicted the next state deterministic instead of probability.

  3. Discrete VS Continuous Environment:

    • Discrete Environment:
      >Environment that has finite number of choices to predict the next state.

    Example: Chess, which has finite number of board positions we can predict the chess moves.

    • Continuous Environment:
      > Environment that has possible inifinite number of position or states that will make the decision process tricky.

    Example: Dart, which has infinite number of states and position of dart.

  4. Benign VS Adversarial Environment: – Weather[Random but not contradict] VS Chess

    • Benign Environment:
      > Environment the states that will not contradict its own rules/conditions byt its outcome result. Which make computing easier.

    Example: Weather, its state can be random and it will not contradict its own state like winter will not change into summer in a hour.

    • Adversarial Environment:
      > Environment the state will contridict the previous or present state of action. Which makes the computing harder.

    Example: Chess, the opponent state cant be the same / predictable all the time.

Hope you understood something out of it, Shout out comments and suggestions in comment.

A journey towards Artificial Intelligence – Introduction

AI is next big thing in computing AKA Transcendence and some says it may end of humanity AKA Terminator, As i getting deep enough it fascinate me and here is few things i like to share, based on udacity course

AI is called as Intelligent Agent, so what is Intelligent Agent?

Consider Intelligent agent as a terminator, that can
1. Able to get data about the environment through sensors, Is person is enemy?
2. Make decisions based on sensor data i.e, Actuators , Kill him ?

This process happen in cycle called Perception-Action Cycle

Intelligent Agent

Application of AI in real world:

  1. Finance – Take decision to buy or sell a stock based on rates & Big data for past market statics.
  2. Gaming – Based on your move the game will be played against you.
  3. Robotics – Any type of robot.
  4. Medicine – Personal diagnostic agent, prefer medicine based on vital signs
  5. Web – Search engines

Next will write about AI terminology