I wonder. Is there an n-dimensional complex like number? That is. Could you have some power rule for data with many columns. The power rule would shift the element data into the ?previous lower dimension if squared. Similarly taking the square root ?could (-1) put the element data into the next higher column.
It could be that the new number is hard to understand. So the idea is to separate the data into harder and harder columns. I mean. Machine learning does not care. I think it could make more sense if it computed data with similar data.
The blog post is meant as inspiration. Reinvent the math wheel.