Author Topic: Support Vector Machines  (Read 201 times)

Raistlin

  • Member
  • **
  • Posts: 238
Support Vector Machines
« on: September 01, 2017, 03:43:14 PM »
So, I've been playing around quite a "bit" <p> with various ideas of late. Found this very nice
explanation on the background and implementations of support vector machines (SVM - supervised machine learning).

It's pretty idiot proof - even for a non-mathematician like myself. The possibilities are truly amazing,
imagine the ability to do near-real time classification (differentiation) of multi-dimensional objects
with high accuracy!

Anyhow - here's the link enjoy the read.  :bgrin:

https://quantmacro.wordpress.com/2016/06/14/support-vector-machines-without-tears-part-1/

K_F

  • Member
  • *****
  • Posts: 1287
  • Anybody out there?
Re: Support Vector Machines
« Reply #1 on: September 01, 2017, 10:53:24 PM »
I've been busy playing (part-time) with AI for the past few years, which is gaining some speed lately.

Gaining knowledge started with the usual books/pdfs and stuff and soon found these ideas 'limiting'.

So I'm now at the idea of what I'd call unrestricted, non-dimensional structures in Masm.
It really just is a bunch of defined Input/Output node numbers and the stuff in between is a mess, with no definite structure that always changes.
I'm more or less ready to program the idea, busy developing the hardware assembly system (tinkering post) and will eventually buy the Stratix10 FPGA to ultimately plonk the masm stuff into VHDL.
That's when the Stratix10 is cheap enough - I can only cry about the prices at the moment.
 :biggrin:
« Last Edit: September 02, 2017, 03:32:46 AM by K_F »
'Sire, Sire!... the peasants are Revolting !!!'
'Yes, they are.. aren't they....'

felipe

  • Member
  • ***
  • Posts: 280
  • I love assembly language programming.
Re: Support Vector Machines
« Reply #2 on: September 02, 2017, 01:22:31 AM »
K_F  that sounds like an interesting project  :icon14:
Felipe.

K_F

  • Member
  • *****
  • Posts: 1287
  • Anybody out there?
Re: Support Vector Machines
« Reply #3 on: September 02, 2017, 03:32:08 AM »
 :biggrin:
The wife thinks I'm nutz... but you know women .. they always looking for visible 'proof' and have zero,zip,nudder understanding of algorithms and computational stuff.
But she always live in hope  :t
'Sire, Sire!... the peasants are Revolting !!!'
'Yes, they are.. aren't they....'

LiaoMi

  • Member
  • **
  • Posts: 132
Re: Support Vector Machines
« Reply #4 on: September 02, 2017, 03:51:11 AM »
I've been busy playing (part-time) with AI for the past few years, which is gaining some speed lately.

Hello K_F!

You have a lot of experience, could you advise public libraries for experiments with artificial intelligence? I could connect them to my project and sketch something interesting. For example, something like this http://blog.dlib.net/2014/02/dlib-186-released-make-your-own-object.html , but closer related with artificial intelligence  :icon_rolleyes:

Best regards

K_F

  • Member
  • *****
  • Posts: 1287
  • Anybody out there?
Re: Support Vector Machines
« Reply #5 on: September 03, 2017, 06:15:51 PM »
If you have a look at Nvidia, AMD, Intel websites, and maybe others, you'll find they have these libraries available.

I've steered clear of already made libraries as they're most likely programmed in C or C++, and probably follow set mathematical algorithms.
Besides not wanting to rely on what others think AI should be, I'm just going ahead and breaking that concept.

I won't be here for a few weeks (well maybe, if my tablet does run overseas) as we're off on holiday/uni registration in a few hours.
Time to complete packing.
 :t
'Sire, Sire!... the peasants are Revolting !!!'
'Yes, they are.. aren't they....'