Perceptron - An Idle Game

Andrew Zuo

Contains adsIn-app purchases

About this game

Perceptron is a new incremental game which is based around the idea of building and training a neural network. The concepts behind a neural network have been distilled to their essence in this idle simulation.

Not to say the game is simple. Sure, it starts simple with only nodes, training, and data, but soon balloons into a complex idle game with prestige and upgrades. Not to mention offline support.

Take on the role of a young undergraduate student as you become quite the idle tycoon. Soon you'll be rivaling even GPT-3.

Perceptron is not just another idle clicker. It was designed around the idea of training a neural network and has taken inspiration from many neural network topics. Who knows, you might even end up learning something.

Ratings and reviews

"I don't know what I'm doing: The Game" It's slow, very slow. But it's a functional idle game? Three stars because no forced ads. Edit: Two months later, still feel a little lost but my appreciation has gone up quite a lot. Hit quite a few automation breakpoints and it's more enjoyable and has become one of my go-to idlers to check in the morning and on breaks.

As a fan of idle games this is particularly... Bland. It's not actively bad, but there's very little depth and prestige, progress is slow by idle standards, and there's not much offered to keep bringing you back. It uses a lot of technical jargon with no explanation, and overall requires far more player interaction than one would expect. Automation is slim, and I found the game to be a chore to knock out once a day after some time.

First the good: It's got potential. The "prestige" actually feels like an upgrade rather than an annoying requirement. Slowness could be a good thing. I like the simplicity. Second the bad: if I don't force stop it every time I close then it eats my battery really fast. I know real networks are hungry, but you should adjust because it's a game. Also, in real life larger networks aren't always better. Suggestions: Regularization. Different kinds of nodes and activation functions.

What's new

- Fixed a bug with incorrect idle income when resuming (not restarting)
- Fixed bug with 0 appearing as '-0.01'
- Upgraded dependencies