Learning ML

Hello, This is Sagar Devkota, a final year student of Software Engineering. I love the fact that we can give instructions to the machine to make it work as our need. I am being doing so for last 4 years. I write program for / with Nodejs, Android , Angular, Unity3D, LibGDX, Godot, Django, Linux and also write some blog. Most of the projects I worked on are either needed for me or I find it useful some people around me. But I used to believe I could not build stuffs with AI and ML that I can use of, Or I dont have that much knowledge to do so.

I have friends who are AI enthusiast. A year back, I had talked about a silly idea, which need AI and ML stuffs, with a friend Madhav. While taking about the final year project, he suggested to work on that idea. Probably this is the time to work on it or forget about it. And I decide to give it a try.

With college, internship and some personal project, I can manage 12 hours a week for this Idea. So the project will take longer time. As I had no plan to learn and work with AI and ML any sooner. Now its changed and with this change, I like to added up another change with it, and it is to keep track of the learning and the project. So I will be sharing any progress on learing and the project here.


The Project Idea

We are inside Computer Simulation

Do you know how we feel hungry? The property that represent hunger goes down. when we eat something the value will be increased and it decreased gradually, then we feel hungry again. The rise and fall depends on the food (exactly the value property) and the activity (the effort property) we do. Same goes for all other.

But I dont believe on this. :)

Let the project prove me wrong.

The Idea:

There will a environment with some objects, the object will have multiple properties like age, hunger, happiness. There will be other objects having different properties, that will fulfill the objects need, they are visible to the object. Then we let the machine learn itself to get the objects around and fulfill the needs. The program keeps on running 24/7. Also we should have some snapshot of every day. Also There should be option to start another instance at any moment, letting both run, we can find how small change effects the process.

If the project looks good and keep going, we can introduce the complexity to the objects.

My current knowledge on this stuffs

I am not totally new on AI, as I had studies 3 credit course named Artificial Intelligence and Neural Networks. I was not that interested back then only things I remember is Knowledge base, prolog, cross over, mutation. just the terms.

What I am talking about (the Idea) is probably what called unsupervised learning.

Timeline

October 13 2020

Started this webpage
Talked with friend, suggested to understand theory on Genetic Algorithm.

- October 17 2020

Genetic Algorithm
Attempt a clone of Smart Rocket [initial].
Looks it need a lots of maths, :(
Neural Network required, as must (similar)projects out there used it.