Forem: Venkata Sugunadithya

Linear Regression: An Overview

Venkata Sugunadithya — Fri, 06 Feb 2026 14:44:03 +0000

Machine Learning is the science of teaching machines to do better than plain, hard-coded programs.

Instead of telling a computer exactly what to do at every step, we let it learn from data. These learning algorithms help machines spot patterns, understand the mathematical relationships behind those patterns, and use them to make predictions on new, unseen inputs.

Sounds interesting, right?
Curious about how a machine actually learns all this? 👀

This guide will walk you through one of the simplest — yet most powerful — concepts in machine learning: Simple Linear Regression.

What is Linear Regression?

Linear Regression is an algorithm that helps us fit a straight line to a given set of data.

Take a look at the image below. You’ll notice several data points scattered across the graph. Technically, you could draw many different lines through these points — but why do we choose this one specific line?
That’s exactly what the machine learning algorithm figures out.

The goal is to find a line that predicts values as close as possible to the actual data points. In simple terms, the algorithm fits the line in such a way that the overall prediction error is minimized — and this condition holds true for all the data points.

Now that we know we’re dealing with a line, what do you think the algorithm is actually trying to predict?

Exactly — it tries to learn a function f(x) that represents a straight line.
And if my Indian folks here have survived 11th and 12th grade math, you already know where this is going 😄
Yep, we’re predicting the equation of a line.

How do we predict this line?

To do that, we need to understand two important things:

The Cost Function
Gradient Descent

These two work together to help the algorithm figure out how good a line is and how to improve it step by step.

In the next part of this guide, we’ll dive into the cost function — where we’ll see how the algorithm measures error and how it tries to fit the best possible line to the data.

More math, more intuition, and less confusion 😉
Stay tuned.

Day 10: Finally a 2 digit number!!!

Venkata Sugunadithya — Thu, 29 Jan 2026 17:32:03 +0000

Hello There Folks!!!
well today was another day of grinding ml....I successfully completed multiple regression and did some dsa questions from codechef, more content after.....

apart from the daily wins....I did something special today....I made my first submisson on kaggle!!! It was this ever running house sale price predictions competetion, got a really bad score with a rank of 4352. The model I used is a simple linear regression that works on a single parameter that is area of lot....so will have to look into other data to and build a better model

What did I do today?

dsa(Number Theory) Doing these questions made me realize constraints are really Importent....of cource I knew this, but today I felt it.

-Kaggle

There you go thats my first submisson I made....You can find the code for the same in the bellow link
Linear Regression using cost function and gradient decent

challo then see you again.....

Day 9: yup we are back, strong as ever.....

Venkata Sugunadithya — Tue, 27 Jan 2026 17:34:16 +0000

Hello There !!
Today was fun, after a long break I decided to work on writeing full code for gradient decent from scratch, and also I will write a full guide for linear regression here on dev soon

This guide will the intution behind cost function and gradient decent along with how learning rate effects the results....all these days of research wont go on scrap, ill share it here so everyone can access, since as I observed, there are very limited resources out there that help in Machine Learning

Well that aside this is what I am building right now....still got things to sort out

Dataset I Used:
Sales price of different houses along with other factors that effect the same

My Code:
implimentation of linear regression with the help of gradient decent and cost function

People keep cretisizeing that I copied code from somewhere....so here are some issues I faced so that you know this is my work no matter what

Overflow at functions compute_cost, gradient_decent as i didnt initalize cost, dm, db to 0.0 and used the power function insted of **
Forgot that alpha existed and ended up getting a weird graph that looks like this for iterations vs cost
I didn't know what skewing initially was, but i was aware of scaleing and ended up getting this which I absolutely hated cause I had a felling this will mess up my gradient decent cause values were too close to eachother

Final Results after I fixed things I felt were wrong:
Graphs:

Cost Vs area of land

ahh so beautiful, Exactly what I wanted a nice scaled data that is centered
Iterations Vs Min cost function for different values of alpha (the step size)
alpha = 0.0001 painfully slow i had to wait more than usual

alpha = 0.001 I got a really lovely graph for this.....I know I can do better but this was good

Lets see an overstep just in case......
alpha = 10

So thats it for today, tmrw see me do a test on my first model after looking for any refinement.....

Day 8: Yup We Lost.....But We Raise

Venkata Sugunadithya — Fri, 16 Jan 2026 05:33:45 +0000

Here is the truth.....as I said in my portfolio website I document even when things are unfinished, and yes thats true cause here I am and haveing done nothing since the past 8 days, the burn out I felt, the weak mindset I had, have won the battle.

but there is a saying here in hindi that goes something like "Har ke jeetne valonko baziger kehete hai" which translates to "A true winner is one who rises after defeat", well I guess this is something like that.....

the above graph shows something importent that I will follow,
the test that comes after the consistancy, I will not lose to it again, I don't have to cause I genuenly love what I do.....and like any other relation I know this is gonna get better too

sooo.....here I am back from the break, working again
My goals for today are:

learn and practise slideing window concept and problems
go break into multiple linear regression
got a small internship to work on n8n, go deal with that
complete my portfolio
document here on dev....

that's it folks i'll catch up today once again.....untill then
poundin, nodin, see ya!!!

Day 7: Untill I Get An Internship At Google

Venkata Sugunadithya — Thu, 08 Jan 2026 17:53:29 +0000

Day 7 started with something important — a one-day break. After a consistent grind, the pause was needed, and honestly, it helped me reset mentally.

Coming back, I practiced some DSA and then decided to test myself by participating in a CodeChef contest. I was able to solve 3 out of 8 problems. Not a perfect score, but a very real checkpoint. Contests expose gaps fast, and that’s exactly why they’re valuable.

Later, I jumped back into my machine learning course, and wow Andrew Ng really knows how to make concepts click. I started learning multiple linear regression and began wrapping my head around vectorization. This concept genuinely feels like magic, the idea that a change in representation can drastically reduce computation time is fascinating. It finally made me appreciate why ML code is written the way it is.

What I worked on today

Took a needed break to reset
Practiced DSA and participated in a CodeChef contest
Started multiple linear regression
Learned the intuition behind vectorization

Progress isn’t always linear, but understanding why things work feels like a big step forward.

Day 6: Untill I Get An Internship At Google

Venkata Sugunadithya — Tue, 06 Jan 2026 18:25:18 +0000

Day 6 felt like things started to connect.

I kicked off data science in Python, getting comfortable with the ecosystem and exploring how Python is used to work with data. Not sure yet where this path will lead, but I’m letting curiosity guide me.

I also spent time doing DSA practice on CodeChef, keeping the habit alive and reinforcing problem-solving alongside theory.

And the highlight of the day — machine learning.
Today, I finally learned my first complete ML algorithm: Linear Regression, along with how cost functions and gradient descent work together to train a model. Seeing the math, intuition, and code align was incredibly satisfying. It finally feels like I’m not just “reading ML,” but actually understanding how a model learns.

What I worked on today

🐍 Started data science with Python

🧩 Practiced DSA problems on CodeChef

🤖 Learned Linear Regression end-to-end (cost function + gradient descent)

One algorithm down many more to come.

Day 5: Untill I Get An Internship At Google

Venkata Sugunadithya — Mon, 05 Jan 2026 17:13:21 +0000

Day 5 was intentionally lighter, but still meaningful.

I explored the intuition behind gradient descent how models iteratively adjust parameters to minimize error, and why step size and direction matter. I’m still forming a complete mental picture, but this felt like an important conceptual bridge from cost functions to actual optimization.

On the building side, I made a few small tweaks to my portfolio website, refining interactions and layout details. Nothing major, but enough to keep the project moving forward.

What I worked on today

ML: Learned the basics of gradient descent
Made minor portfolio updates

hmmmm....what did i do today?

Some days are about depth, others are about maintenance. Both count.

Day 4: Untill I Get An Internship At Google

Venkata Sugunadithya — Sun, 04 Jan 2026 17:42:46 +0000

Today felt really good.

I went deeper into cost functions and finally built a clearer mental model of what they represent and why minimizing them actually improves a model. I also made small input-level changes in the linear regression model I’m building, which helped connect the math to actual code behavior.

On the DSA side, I revised the two-pointer technique and ended up solving a sliding window problem without even realizing it was two pointers at first. That moment when the solution naturally fits a pattern is super motivating — and yes, the two-pointer approach worked perfectly.

I also shipped a UI interaction update to my portfolio: added a second-page component where clicking the menu causes the main screen to slide and reveal navigation. Subtle, clean, and honestly… it looks really cool.

What I worked on today

ML: Built deeper intuition around cost functions, Updated a linear regression model with input variations
DSA: Revised two pointers and solved a sliding window problem
Portfolio: Added a menu slide interaction to my portfolio website

hmmmm....what did i do today?

What this reminded me
Progress isn’t always loud. Sometimes it’s realizing you already understand more than you thought.

Day 3: Untill I Get An Internship At Google

Venkata Sugunadithya — Sat, 03 Jan 2026 17:09:54 +0000

Day 3 was about slowing down and understanding the math behind ml and implementing it.....

I focused on implementing linear regression from scratch using mathematical intuition rather than jumping straight into libraries. I explored how predictions are formed, how parameters affect the line, and started digging into cost functions — I get the idea at a high level, but I still need to deeply understand what exactly they represent and why they work. That’s on tomorrow’s list.

Alongside ML, I pushed an improvement to my portfolio on GitHub, refining layout and structure as I continue building it incrementally.

I didn’t get to DSA planning today as intended, but instead of forcing it, I’m scheduling it intentionally for tomorrow. Learning consistency > pretending everything went perfectly.

What I worked on today:

ML: Implemented linear regression using math (I studied about regression as a part of college math, it helped me implement this code) and Explored the intuition behind cost functions

Wanna check the code out?
Portfolio: insted of click added a scroll property so that the transition can be made even smooth. more coming soon!!!

Progress Log

What’s next?

Properly break down cost functions
Plan and start DSA practice
Keep building — even if things feel unfinished

Learning in public. One step at a time.

Day 2: Untill I Get An Internship At Google

Venkata Sugunadithya — Fri, 02 Jan 2026 16:24:01 +0000

Today, I explored the foundations of Machine Learning, focusing on the difference between supervised and unsupervised learning. Understanding when and why each approach is used helped clear up a lot of confusion I had earlier around ML being just “models and math.”

I also got into the basics of Linear Regression — what the model is trying to learn, how it fits data, and why it’s often the starting point for supervised learning. I’m deliberately keeping things simple right now and focusing on intuition before diving deeper.

On the building side, I documented my third meaningful push to GitHub for my portfolio website. The project is still evolving, but pushing code felt important — not to show perfection, but to start building a visible trail of progress.

What I worked on today:

ML: Supervised vs Unsupervised Learning and Linear Regression fundamentals
Portfolio: UI updates & structure updates

commit history

This phase is all about understanding before optimizing and building in public without waiting to feel “ready”.
Back at it tomorrow.

Execute Day 1: Untill I Get An Internship At Google

Venkata Sugunadithya — Thu, 01 Jan 2026 17:19:33 +0000

Two Pointers, Machine Learning & Building My Portfolio
committing to showing up — even when things feel slow.

Today, I started working on Data Structures & Algorithms, beginning with the Two Pointers pattern. Instead of rushing through problems, I focused on understanding why the pattern works and when to recognize it. The goal right now isn’t speed — it’s clarity.

Alongside DSA, I also took my first steps into Machine Learning, getting familiar with the fundamentals and what supervised learning actually means beyond the buzzwords. It’s early, but I’m excited about where this path can lead.

On the creative side, I pushed Day 2 progress on my portfolio website. I’m experimenting with layout, typography, and structure — trying to make it reflect how I think and build, not just what I know. It’s still a work in progress, but documenting the journey feels more important than waiting for perfection.

What I worked on today:

Two Pointers pattern (concept)
where I learned it from
ML: Getting started with the basics
Portfolio: Layout refinements & structure updates

commit history

The plan is simple:
learn → build → document → repeat.

More tomorrow.