Ask HN: Who has sold an algorithm and how? | Hacker News Reader

Ask HN: Who has sold an algorithm and how?

21 pointsdaniel-cussen12 comments

Can anyone point me to a person that has successfully sold an algorithm, a general algorithm, for an appropriate amount of money? How did the transaction take place? What would an algorithm writer need to do before the transaction? Take PageRank. Larry and Sergey were unable to sell it as an algorithm or even as a website. What should they have done, short of starting Google, to get a fraction of the proceeds for a (surely different) fraction of the work?

View on HN

In some cases, you can patent a major application for that algorithm and then license that patent. From what I know Larry and Sergey patented PageRank for Web Search. My advisor has patent on a Graph Cuts algorithm for image segmentation, which is licensed by Cornell. If you have a fast matrix multiplication algorithm e.g. for use in Fully Connected layers for Neural Nets. You can patent that application. When FB hired people from NYU to work on deep learning, the acquisition involved getting a fast fourier transform optimized version of convolutional layers. Not sure how big of a component it was.

OP here. So one algo I have is fast matrix multiplication algorithm, especially for sparse matrices. Faster asymptotically and in real-time than what's out there, and with better accuracy.

If you really have a matrix mulitplication algorithm that multiplies nxn matrices in less than O(n^2.3729), you have made a mathematical breakthrough and should publish.

I second "publish".

If you've invented something worth noting, you'll get infinitely better residual value, visibility, respect and goodness coming your way for years to come vs. one time deal of selling your work to a private buyer for a one time fee in a sell-and-forget type of deal.

What do you mean by "with better accuracy"? How can one matrix multiplication algorithm be more accurate than another?

Real computers are finite precision. There are very real issues with numerical stability where you are running along fine for many many cases, and then there is a catastrophic cancellation that wipes out all of your accuracy. There is a huge difference between writing a paper on bounds of matrix multiplication, and delivering performance in a finite precision world.

Strassen makes your rounding errors bigger the more levels of recursion you do. We're talking ~7 bits of error due to rounding (in a paper I read).

Hedge fund algorithm developer here. I've sold a few of these but it's mainly been based on taking 20% of the net profit and largely confined to the financial industry.

Just emailed you.

I'm interested in some financial programming. Does everyone use Haskell or other functional languages? Where can I learn some basics?

How about this?

https://www.coursera.org/learn/computational-investing

A more accessible way to do this for trading algorithms would be through Quantopian [1], where they have a crowd-sourced hedge fund that has a similar profit sharing model. You have to have good standing in their contests to be eligible though.

Of course, you can also just use their platform to trade your algorithm with your own money.

[1] https://www.quantopian.com/

Reply on news.ycombinator.com