SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
tagoma
Topic Author
Posts: 18233
Joined: February 21st, 2010, 12:58 pm

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 10:14 am

 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 10:28 am

tagoma wrote:

Nice. Did you but the book?
 
User avatar
tagoma
Topic Author
Posts: 18233
Joined: February 21st, 2010, 12:58 pm

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 10:33 am

outrun wrote:
tagoma wrote:

Nice. Did you but the book?

Book shall be available in February, I read. It was advertised on LI by the author.
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 11:29 am

tagoma wrote:
outrun wrote:
tagoma wrote:

Nice. Did you but the book?

Book shall be available in February, I read. It was advertised on LI by the author.

Ah,..  I did a quick search and saw it at Wiley, but I didn't realize it wasn't available yet. It looks like it has a fund focus, and he has lots of experience and reputation. I'm not sure if it covers the concepts of to the more out-of-the-box approaches that *you're* doing is with ML in finance.

"Marcos has an Erdős #2" .. very funny that it's in his bio, but it indeed caught my attention. I still have this feeling that it must be possible to get an Erdős #1 if you try hard enough.
 
User avatar
Cuchulainn
Posts: 56246
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 1:47 pm

It's too late to ask PE himself?

It's all a bit affectatious, that nonsense. 
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 2:01 pm

Cuchulainn wrote:
It's too late to ask PE himself?

It's all a bit affectatious, that nonsense. 

What's your blood type?
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Machine learning for fun, to pay your rent or whatever

January 23rd, 2018, 2:09 pm

dup
 
User avatar
tagoma
Topic Author
Posts: 18233
Joined: February 21st, 2010, 12:58 pm

Re: Machine learning for fun, to pay your rent or whatever

January 26th, 2018, 11:34 am

In the following video around 4'40'', I don't get well why they separate so firmly loss function and cost function.
Are definitions of these two kinda different in the context of RBM and AU?
Shall I stick to the thin(er) line that is drawn here, instead? 

AUs use backprop in their learning process. Instead of cost, the metric used to assess the quality of the network is loss, which is the amount of information lost in the reconstruction of the input

 
User avatar
Cuchulainn
Posts: 56246
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Machine learning for fun, to pay your rent or whatever

January 26th, 2018, 6:37 pm

Brigo sees machine learning and data science having a big impact in finance, but is sceptical of the near-term prospects for artificial intelligence: “We are still very far from a real intelligence and human-like thinking,” he says. Rather, voice recognition, chess and other game-playing applications, or interactive software that is able to answer questions, are simply examples of sophisticated statistical algorithms.

Brigo asks the question why?

https://www.risk.net/comment/5389041/po ... g-and-more
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Machine learning for fun, to pay your rent or whatever

January 26th, 2018, 6:44 pm

Cuchulainn wrote:
Brigo sees machine learning and data science having a big impact in finance, but is sceptical of the near-term prospects for artificial intelligence: “We are still very far from a real intelligence and human-like thinking,” he says. Rather, voice recognition, chess and other game-playing applications, or interactive software that is able to answer questions, are simply examples of sophisticated statistical algorithms.

Brigo asks the question why?

https://www.risk.net/comment/5389041/po ... g-and-more

The human brain is merely another example of a sophisticated statistical algorithm. At this point, all we are arguing about is the level of sophistication.
 
User avatar
ISayMoo
Posts: 819
Joined: September 30th, 2015, 8:30 pm

Re: Machine learning for fun, to pay your rent or whatever

January 27th, 2018, 12:41 am

I would argue that learning to play Go or Chess from scratch, without looking at expert play, just from self play, goes beyond what most people understand by "statistical algorithms".
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Machine learning for fun, to pay your rent or whatever

January 27th, 2018, 3:56 pm

Exactly!

What is amusing is that every time ML/AI achieves a goal (beating the best humans at chess, go, poker, etc. or translating foreign languages, recognizing images, diagnosing disease, etc. ) that was declared to mark the boundary between human and machine intelligence, the human-egoists move the goal posts.
 
User avatar
Cuchulainn
Posts: 56246
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Machine learning for fun, to pay your rent or whatever

January 27th, 2018, 5:47 pm

Traden4Alpha wrote:
Exactly!

What is amusing is that every time ML/AI achieves a goal (beating the best humans at chess, go, poker, etc. or translating foreign languages, recognizing images, diagnosing disease, etc. ) that was declared to mark the boundary between human and machine intelligence, the human-egoists move the goal
posts.

Infidels!

The etc. part might fuel the Doubting Thomas.  They think its's Expert Systems revisited. An important question is which problems are NOT suitable for AI? No one seems to own up on that one.

https://en.wikipedia.org/wiki/Doubting_Thomas
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Machine learning for fun, to pay your rent or whatever

January 27th, 2018, 9:27 pm

Cuchulainn wrote:
Traden4Alpha wrote:
Exactly!

What is amusing is that every time ML/AI achieves a goal (beating the best humans at chess, go, poker, etc. or translating foreign languages, recognizing images, diagnosing disease, etc. ) that was declared to mark the boundary between human and machine intelligence, the human-egoists move the goal
posts.

Infidels!

The etc. part might fuel the Doubting Thomas.  They think its's Expert Systems revisited. An important question is which problems are NOT suitable for AI? No one seems to own up on that one.

https://en.wikipedia.org/wiki/Doubting_Thomas
The more important question is the structure of the Venn diagram between problems that are NOT suitable for humans and problems that are NOT suitable for AIs.
 
User avatar
ISayMoo
Posts: 819
Joined: September 30th, 2015, 8:30 pm

Re: Machine learning for fun, to pay your rent or whatever

January 28th, 2018, 12:38 am

Depends if you mean AI = smart algo, or AGI (Artificial General Intelligence) Any problem suitable for humans will be suitable for an AGI, by definition.
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...