Introduction to Machine Learning
What is machine learning?
Lets go to the origin of machine learning , the term “machine learning ” was coined by Arthur Samuel (an IBM employee).
Now why machine learning , why do we need machine learning in this world in the first place ?
Well machine learning is the most promising field of computer science
And it does delivers the promising aspects of modern computational architecture and the infrastructure for the ease of day to day task and on going projects , but do we really need machine learning for our aspirations in the computer science field why not opt for java ,c++, web dev and other language and field why machine learning is the only thing going on right now why everyone betting on it and other tech geeks and enthusiast are advicing to start integrating this particular thing in your workflow .
Why machine learning and why it should be.
- Machine learning provides automation which is absolutely vital for productivity and ease of your workflow .
- Machine learning is what can build the AI tools which is absolutely going off the roof this decade and might be the upcoming future where all the automation and customer service will be provided by it .
- It integrates with every language name , webdev(html ,css , js), mern stack, java ,c ,etc..
- The machine learning requires python which relatively the most easily comprehended language to code with so you get the best tool with minimal code efforts
- Machine learning can ease the work of analysts , data scientists , traders,prediction gamblers , etc..
So why not this should be your attitude to the machine learning not be rebellious towards it rather embrace it in your workflow to get that off the stress that your predecessors had and had to go through it at least you can work now with assistant and be free to think and wander to your ideas that was holding you back .
Let’s head back to the topic again: what is machine learning exactly? We figured out why and why we should use it now lets understand how it is .
Machine learning is the use of computational algorithms that learn from data to make predictions or decisions without being explicitly programmed.
In the nutshell , think of this like you provide some context to it and it generates the solution based on your query and with that context it can gain experience to solve your problems.
I am not using too much lingoes(like supervised and unsupervised learning ) .In the further blogs we will use some Technical terminologies to connect with the theoretical part of the machine learning and ai too, this blogs relatively practical and uses layman language to reader so it doesn’t feel overwhelming at all .
How is machine learning connected to Artificial intelligence i.e ai ?
as you can see in this picture the relation between the two don’t get into the deep learning section we will cover it soon. Let’s come back to the topic and discuss the working and differences of the two .
Artificial intelligence
the buzzword of the 2020s is the one and only is ai i.e artificial intelligence how it works. Hmm, let’s take a look .
We are gonna use some terms here , let’s dig into it
- Data processing ( process and integration of the data in the ai’s memory primarily the model it’s trained in ).
- Data preprocessing(although it seems like the above it but is completely different from it the former integrates the data and the latter cleans the data and processes to make sense to the machine and algorithm).
- Vector embedding(although i will cover this topic in upcoming blogs so let’s just see this in a short , the vector is simply the connection between one to another data points to make relation between them(numerically) and use it in a gathered context dont worry if it seems complex it will be completely clarified in upcoming blogs(no need to worry).
- Use of LLms to process the input and work on the output via various processing layers such as attention, here we can see in the picture let’s simply it out ,
The attention layer in simple words basically does the job of processing the input layers and by the use of Natural language Processing(NLP)
It uses the embedding process and vectorization to process ahead to the feed-forward block which simply means that it predicts the next most probable token/word.
Lastly the softmax in the attention block proceeds the numeric encoded vectors to the probability and gives the value of 1 which is the highest.
This much explanation would be adequate for now although we will discuss them in detail in an upcoming blog so stay tuned.
You might be thinking this all thing mentioned here is about LLm not ai well you are absolutely wrong the bedrock of ai is LLm itself without it can never function like it does there’s no point of having an ai which cannot be trained and fine-tuned
Further.
So without further ado let’s deep dive into the machine learning complex jargons and simplify them off so next time when you are in a room full of tech geeks you dont afraid to share your opinion as the purpose of this blog post is exactly that educating and making yourself relevant in the world of relentless changing technologies via blogs.
You might be confused that whats is encoding and decode exactly
- encoding
The process of converting the text into the numerical value for the vectorization in order to make embeddings so it will be easier for the machine to compile the inputs and put them in a relational arrangement
- decoding
The opposite of encoding as the name suggest , decoding which just converts the numerical representation of the vectors into text or words to perhaps generate the output
- Positional encoding
Positional encoding is again the arrangement of the vector towards the position where it all makes sense as the context is provided to the LLm.
Five things you need to get started on machine learning quest:-
- Make yourself obsessed with machine learning and its application: read , see and visualise the stuff about machine learning in your daily life .
- Let your mind wander so you can build something of your own , in this way the project will incorporate your own thinking and intelligence. It would be eclectic to see your thinking translated via code to interface. Trust me when you do it , you will know it .
- Machine learning isn’t everyone’s favorite until they start making projects as this is largely backend this doesn’t provide short term gratification you need to stick around to actually witness the power of machine learning
- Start with smaller and simpler projects like , iris classification , house price prediction this will strengthen the core basics of machine learning as its the utmost priority if you want to master machine learning .
- Be patient. The most important quality you ever want is to be patient. Sometimes your model will make mistakes if you abandon it and go to the next one. You will waste time and resilience both so be patient when building projects take the time to understand each and every line of the code. No line should seem magical or strange.
Bottom line Machine learning is the present and is the most probable future too
Get started with it even though you feel just 30 mins a day it will take you ahead from where you are, in the next blog we will discuss the types of algorithms used in machine learning and what’s the advantage and disadvantages of that