AI,ML,DL-Overview:Keeping it simple...

01. How are they related ?

AI is the super set. ML is the sub set of AI. DL is the subset of ML.

02. AI (Artificial Intelligence):

a.      Artificial Narrow Intelligence (ANI):

1. AI that we are using now
2. AI which is not equipped without Self Awareness, Emotional Intelligence
3. AI which still needs Human assistance, towards it’s working

b.      Artificial General Intelligence (AGI):

1. AI that we are not using now
2.  AI which will be equipped with Self Awareness, Emotional Intelligence
3. AI which will require minimal Human assistance, towards it’s working

c.      Artificial Super Intelligence (ASI):

1. AI that we are not using now
2. AI which will be equipped with Self Awareness, Emotional & all Intelligence which Humans do
3. AI which will require no Human assistance, towards it’s working and it will supersede Humans in all aspects

03. ML (Machine Learning):

a.      ML usage: It’s used for processing of Limited amount of Data; Simple & Medium scale Problems.

b.      For ML Model(s) to work: Model(s) work based on their training with Features of Object(s) called Training Data, and with help of Algorithm(s).

c.      Supporting Elements in getting a ML Model ready:

1. Training Data (Features of Object(s)) (involves Data Gathering, Cleaning, Exploratory Data Analysis)
2. Algorithm
3. Model = Training Data + Algorithm
4. Predictor Variable(s)
5. Response Variable(s)
6. Model Testing Data

d.      Type of Problems, ML solves:

1. Regression: It's also called 'Continuous Quantity'. Objective: To Forecast/ Predict. E.g: Stock Price. Algorithm used: Linear
2. Classification: It's also called 'Categorical Quantity'. Objective: To Categorize. E.g: Emails into Spam or Non-Spam folders in Gmail. Algorithms used: Logistic, Decision Tree, Random Forest (Multiple Decision Trees), Naive Bayes (Based on Objects and it's Features), KNN (Nearest Neighbor), SVM (Support Vector Machines)
3. Clustering: Objective: To Group. E.g: Grouping of transactions which are fraudulent in nature. Algorithm used: K-Means (where K means no. of Clusters/ Groups we need to form)

e.      Types of ML:

1. Supervised Learning: Used for solving Regression & Classification Problems. Here the Model is trained using Labelled Data Sets.
2. Unsupervised Learning: Used for solving Clustering Problem. Here the Model trains itself by studying Unlabeled Data Sets & Self Grouping.
3. Reinforcement Learning: Here an Agent is placed in an Environment, which learns Data by itself – by performing Actions and discovers Errors or Rewards

04. DL (Deep Learning):

a.      DL usage: It’s used for processing of Huge Amount of & High Dimensional Data; Complex Problems & Image, Handwriting, Object Recognition.

b.      For DL Model(s) to work: Model(s) work based on automatically created Features of Object(s), and with help of Algorithm(s).

c.      Types of DL/ ANN (Artificial Neural Network):

1. Single Layer - Node/ Neuron/ Perceptron, which has following Elements:

• Input Signals (X1, X2, Xn...) with their respective Value/ Weights (W1, W2, Wn...), sends it's input to Weighted Sum.
• Weighted Sum = X1*W1+X2*W2+Xn*Wn, sends it's input to Activation Function.
• Activation Function, finally calculates the Output Value

2. Multi Layer - Node/ Neuron/ Perceptron: In here, we have multiple Nodes like above - distributed across Inputs Layers, Hidden Layers & Output Lawyers. Wherein out of Nodes from Input Layer travels thru Nodes in Hidden Layer to finally the Nodes in Output Layers -- which then generate the End Result/ Final Output. Note:: We can have single or Multiple Hidden Layers. E.g below:

Following are type of Multi Layer - Nodes/ Neuron/ Perceptron:

• FFN (Feed Forward Network): In here, Previous Output has no relation to the Next Output.
• FBN (Feed Back/ Back propagation Network): In here, we use the Gradient Descent Method, by adjusting the Weights several times/ iterations - to minimize Error/ Loss and to come to the desired Output. E.g below:

• RNN (Recurrent Neuron Network): In here, Previous Output decides Next Output. Which is exactly opposite to FFN (Feed Forward Network).

This concludes a quick overview of AI, Ml, DL.

Hope you liked it. Do share your comments...

Actually - Each topic above, is a vast ocean in itself and there are numerous Books, Articles, Blogs, Posts - being written on it almost daily - and they are interesting to read & learn ! Man kind is still exploring these areas daily to see it's true potential and coming up with new Algorithms & Models. But indeed, it's a Game Changer and is going to drastically Transform the World & the way Humans live their Daily Routine - in years to come.

So let's stay tuned and keep Exploring, keep Learning :).