Skip to content
← Academy

How a network learns · CHAPTER 2 OF 7

~6 min read ·

The reward defines what “good” means

A neural network understands neither money nor markets. It understands one thing: after every action it receives a number, and it tries to make those numbers as good as possible. We call that number the reward, and it is the only sentence in which we tell a kverk what we want. Everything else — senses, brain size, fuses — is circumstance.

It is also the single most important decision in the system. A badly phrased reward cannot be rescued by a perfect architecture.

The rest is free — you just need to sign in

Course progress belongs to your account, not to your browser — a half-read chapter waits for you on the laptop and on the phone alike. Without an account nothing is saved.

Join the waiting list a code by e-mail · no password

The first chapter of every course is public. Chapters 2+ and the quizzes need a sign-in — still free.

Underlined terms lead to their glossary entry — a definition, an explanation and what the term means in KVERX. More in the glossary →

The academy is written and maintained by Caroline Technologies, the company behind KVERX. What it says about KVERX follows the product’s own rules — see How it works. Nothing here is investment advice: Risk warning.