Is NumPy taught here?

Enough of it to do the project — array creation, shapes, matrix multiplication and the SVD call. It is not a NumPy course; the emphasis stays on knowing what operation you want before you go looking for the function that does it.

How does this relate to PCA?

PCA is essentially SVD applied to centred data, so this sprint gives you the whole mechanism. The dimensionality reduction sprint uses it on real datasets, and the Classical Machine Learning phase course treats it as a modelling tool.

I did linear algebra at university and hated it. Will this be different?

Probably, because the order is reversed. Nothing is defined before you have seen it do something, and there are no proofs at all. Several ideas that are usually introduced abstractly — eigenvectors especially — arrive here as answers to a question you have already been asked to struggle with.

Build sprints₹399

Linear algebra for AI

Take this if you have been told linear algebra is essential for machine learning and cannot see why row reduction would be.

What you’ll be able to do

You can read a matrix as a transformation rather than a grid of numbers, explain what an eigenvector is geometrically, and use SVD to compress real data with a stated quality trade-off.

What it assumes

Comfort with coordinates and basic algebra. Some Python helps for the project but the syntax needed is introduced.

Price: ₹399
Length: 4 weeks
Sessions: 12
Effort: ~5 h/week · 20 h total
For: Class 11–12 · College · Working professional

Adds it to My courses on your dashboard so you can find it again. Payment is not open yet — nothing is charged.

What’s taught, in order

01
Vectors as data, not arrows
A row of a spreadsheet is a vector; a dot product is a similarity score. Starting from data rather than from physics is why this sprint sticks where a first linear algebra course often does not.
02
A matrix is a thing that does something
Every matrix you meet is applied to a shape and watched, so rotation, shear and projection become things you have seen happen rather than definitions. Multiplication becomes composition of two moves.
03
Eigenvectors: the directions that survive
The directions a transformation only stretches, found by hand on two-by-two examples and then read off the visualiser. This is the concept PCA is built on, so it earns real time.
04
SVD, and compression you can see
Decomposing an image and rebuilding it from progressively fewer components, watching detail arrive in order of importance. Abstract decomposition becomes an obviously useful operation.

What you build

Compress an image of your own with SVD at four different ranks. Produce a short write-up with the reconstructions side by side, the storage each one needs, and your judgement of the lowest rank that is still acceptable — plus the reason a photograph of text behaves differently from a photograph of a face.

Check it yourself against this

Four reconstructions at clearly different ranks, with the rank labelled on each.
Storage cost computed for each rank, not estimated.
A stated choice of acceptable rank with a reason that refers to what is lost.
An explanation of which kinds of image compress well under SVD and which do not.
You can say what the singular values are ordering, and why that order matters.

Then answer this

SVD keeps the largest singular values first — what does a large singular value actually mean about the data?

Two or three sentences, in your own words. If you cannot, the course is not finished — go back to the module it came from.

What this course does not do

This is applied linear algebra: there are no proofs, no general vector space axioms, no Jordan forms and no numerical stability analysis. If you need a formal first course for a university exam, this will help your intuition but will not cover your syllabus.

Where this leads

The SVD visualiser used in the last week of this sprint is free on its own page, and pushing the rank slider down until a face stops being recognisable is the single clearest thirty seconds in the whole subject.

Try the SVD visualiser Try a real lesson free

FAQ

Is NumPy taught here?: Enough of it to do the project — array creation, shapes, matrix multiplication and the SVD call. It is not a NumPy course; the emphasis stays on knowing what operation you want before you go looking for the function that does it.
How does this relate to PCA?: PCA is essentially SVD applied to centred data, so this sprint gives you the whole mechanism. The dimensionality reduction sprint uses it on real datasets, and the Classical Machine Learning phase course treats it as a modelling tool.
I did linear algebra at university and hated it. Will this be different?: Probably, because the order is reversed. Nothing is defined before you have seen it do something, and there are no proofs at all. Several ideas that are usually introduced abstractly — eigenvectors especially — arrive here as answers to a question you have already been asked to struggle with.

Other courses

This is a standalone course. It does not add weeks to the 78-week curriculum or change your roadmap. If you want the whole path instead, the full programme is ₹399/month · ₹2,000/year — see what it covers.