Skip to content
AriadneTechnology

The centre · Final exam

The Riddle of the Sphinx

She guards the road to the inner labyrinth with riddles written in symbols. Three wrong answers and you're turned back.

15 min 300 XP bountyBoss battle

In this chamber you will

  • Prove you can read the language of machine learning
  • Earn your certificate of completion

How the battle works

The Sphinx's riddles come from all nine chambers: Greek letters and indices, sets and logic, functions, sums and products, shapes, derivatives, probability, and the equations of real papers. Every riddle is new, written for this battle alone. Some ask you to read a line of notation, some to reason about it, and many to calculate, so keep a pen or a Python prompt nearby.

  • Land 8 correct answers to make the Sphinx yield.
  • Every wrong answer costs one of your 3 lives. Lose them all and you're turned back, but you can always return.
  • There are no retries within a question. Read the symbols carefully, then answer.
  • Victory earns a large XP bounty, the Riddle Solver badge, the legendary Riddle of the Sphinx card and your certificate of completion. Win without losing a single life for a secret extra reward.

After the labyrinth

Look back at how far you've come. In the first chamber, softmax(QKTdk)V\mathrm{softmax}\big(\frac{QK^T}{\sqrt{d_k}}\big)V was a row of hieroglyphs. Since then you've learned to say every symbol aloud and to read sets, quantifiers, functions, sums, shapes, gradients and probabilities. You've followed and written proofs, decoded equations from real papers (attention, Adam, dropout, the GAN objective, the ELBO), and turned several of them into working code. That's the skill this course promised: you can read the language machine learning is written in.

The next step is to understand what that language describes. Linear Algebra for Machine Learning picks up where Chamber 6 left off. It turns vectors and matrices from notation into geometry: spans, transformations, projections, eigenvectors and the singular value decomposition, each tied back to a model you'll recognise. You already know how to read its notation, so it can spend its time on the ideas.