MarcoSteinke / CollatzML

Basic repository to try different datasets and programmatically control Chart.js

📈 CollatzML:

👋 1. Introduction:

Using Collatz Conjectures, ML5 and Chart.js to encode stocks prices and try to use them to predict stock prices for a very short interval of time and with a low precision.

Future features:

• ❏ methods for scaling, stretching datasets, concatenating (2.2)

• ❏ MSE function to measure distance of a Collatz series for a specific n to a certain stock (2.3)

• ❏ neural network which is able to approximate a series of factors with length of the associated Collatz series to reduce MSE and model a stock (3.1)

• ❏ maybe using some tensorflow to model stocks (3.2)

2. Implementation:

In this section I am going to present the proccess of implementing the certain features from "Future Features".

2.1 Methods for scaling, stretching datasets:

In this project’s context a dataset is a simple Collatz series. I want to offer some methods which can be used to manipulate the Collatz series as a whole. Thus the methods will allow the user to scale, stretch and add Collatz series.

Collatz.add(series):

This method will be used to add two Collatz series element-wise. The implementation is very simple:

add(coll) {
  return this.result.map((datapoint, index) => datapoint + coll[index]);
}

It simply maps the first series to the sum of the picked element and the element at the same index in the other series element-wise.

Usage:

  const coll = new Collatz(41)
  // Generate array of integers up to the length of coll
  const integers = [...Array(coll.getResult().length).entries()].map(e => e[0])

  // finally add the two series:
  const resultSeries = coll.add(integers);

⚖ License:

TBD