Serving the Quantitative Finance Community

 
User avatar
outrun
Posts: 4573
Joined: January 1st, 1970, 12:00 am

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 5th, 2018, 1:54 pm

Hello, I'm after the detailed agenda of this event (Nov.30 and Feb.22, 2 dates same agenda?).
The link at the top of this page is broken. That other link on LI isn't so relevant.
Can anyone help, please? Merci.
Hi Tagoma, some things changed between the two dates, we replaced some topics with new ones, the overall theme was the same of course. Are you planning to attend these past events using the time machine below your house?
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 5th, 2018, 3:28 pm

Hello, I'm after the detailed agenda of this event (Nov.30 and Feb.22, 2 dates same agenda?).
The link at the top of this page is broken. That other link on LI isn't so relevant.
Can anyone help, please? Merci.
Hi Tagoma, some things changed between the two dates, we replaced some topics with new ones, the overall theme was the same of course. Are you planning to attend these past events using the time machine below your house?
Hello outrun!
Turning the LHC into a time machine is tempting, but my actual plan is less ambitious!
In fact, I'm searching for research topics mixing finance with ML lato sensu (no Venn diagram provided, though) and heavily commodities-oriented.
I thought maybe the agendas of these 2 recent conferences would bring me ideas. Unfortunately, I can't find them online (anymore)!
 
User avatar
outrun
Posts: 4573
Joined: January 1st, 1970, 12:00 am

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 5th, 2018, 4:21 pm

Hello, I'm after the detailed agenda of this event (Nov.30 and Feb.22, 2 dates same agenda?).
The link at the top of this page is broken. That other link on LI isn't so relevant.
Can anyone help, please? Merci.
Hi Tagoma, some things changed between the two dates, we replaced some topics with new ones, the overall theme was the same of course. Are you planning to attend these past events using the time machine below your house?
Hello outrun!
Turning the LHC into a time machine is tempting, but my actual plan is less ambitious!
In fact, I'm searching for research topics mixing finance with ML lato sensu (no Venn diagram provided, though) and heavily commodities-oriented.
I thought maybe the agendas of these 2 recent conferences would bring me ideas. Unfortunately, I can't find them online (anymore)!
Ah, I see!
Some things we presented were: modelling the dynamic behavior of time series, .. how to model the memory and also the shapes of the return distribution. Also memory free SDE type of models. Historical & implied vol dynamics. Non-linear dimension reduction like PCA -also for groups of commodities-, pairs trading, reconstructing missing data, intraday return & volume dynamics. Outlier detection. Denoising. Also some generic concepts like GAN, modelling (conditional) high dimensional probability distributions, clustering, making sure things don't blow up, simulating multi-variate timeseries with non-Gaussian distributions, speeding up pricing models with function approximation. ..so many things, I'm sure I forgot at least half of it! I have to look at our notes.

Are you planning to do research for fun, for your company, for the sake of science?
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 5th, 2018, 6:13 pm

Hi Tagoma, some things changed between the two dates, we replaced some topics with new ones, the overall theme was the same of course. Are you planning to attend these past events using the time machine below your house?
Hello outrun!
Turning the LHC into a time machine is tempting, but my actual plan is less ambitious!
In fact, I'm searching for research topics mixing finance with ML lato sensu (no Venn diagram provided, though) and heavily commodities-oriented.
I thought maybe the agendas of these 2 recent conferences would bring me ideas. Unfortunately, I can't find them online (anymore)!
Ah, I see!
Some things we presented were: modelling the dynamic behavior of time series, .. how to model the memory and also the shapes of the return distribution. Also memory free SDE type of models. Historical & implied vol dynamics. Non-linear dimension reduction like PCA -also for groups of commodities-, pairs trading, reconstructing missing data, intraday return & volume dynamics. Outlier detection. Denoising. Also some generic concepts like GAN, modelling (conditional) high dimensional probability distributions, clustering, making sure things don't blow up, simulating multi-variate timeseries with non-Gaussian distributions, speeding up pricing models with function approximation. ..so many things, I'm sure I forgot at least half of it! I have to look at our notes.

Are you planning to do research for fun, for your company, for the sake of science?
Thank you for these research topic ideas. It would (mostly) be researching at home, but related to markets I work on.
 
User avatar
outrun
Posts: 4573
Joined: January 1st, 1970, 12:00 am

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 5th, 2018, 6:59 pm

Hello outrun!
Turning the LHC into a time machine is tempting, but my actual plan is less ambitious!
In fact, I'm searching for research topics mixing finance with ML lato sensu (no Venn diagram provided, though) and heavily commodities-oriented.
I thought maybe the agendas of these 2 recent conferences would bring me ideas. Unfortunately, I can't find them online (anymore)!
Ah, I see!
Some things we presented were: modelling the dynamic behavior of time series, .. how to model the memory and also the shapes of the return distribution. Also memory free SDE type of models. Historical & implied vol dynamics. Non-linear dimension reduction like PCA -also for groups of commodities-, pairs trading, reconstructing missing data, intraday return & volume dynamics. Outlier detection. Denoising. Also some generic concepts like GAN, modelling (conditional) high dimensional probability distributions, clustering, making sure things don't blow up, simulating multi-variate timeseries with non-Gaussian distributions, speeding up pricing models with function approximation. ..so many things, I'm sure I forgot at least half of it! I have to look at our notes.

Are you planning to do research for fun, for your company, for the sake of science?
Thank you for these research topic ideas. It would (mostly) be researching at home, but related to markets I work on.
You had some very cool ideas in the past. Commodity markets have so many interesting domain details, I can't wait to hear what your next project will be!
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 5th, 2018, 8:33 pm

Ah, I see!
Some things we presented were: modelling the dynamic behavior of time series, .. how to model the memory and also the shapes of the return distribution. Also memory free SDE type of models. Historical & implied vol dynamics. Non-linear dimension reduction like PCA -also for groups of commodities-, pairs trading, reconstructing missing data, intraday return & volume dynamics. Outlier detection. Denoising. Also some generic concepts like GAN, modelling (conditional) high dimensional probability distributions, clustering, making sure things don't blow up, simulating multi-variate timeseries with non-Gaussian distributions, speeding up pricing models with function approximation. ..so many things, I'm sure I forgot at least half of it! I have to look at our notes.

Are you planning to do research for fun, for your company, for the sake of science?
Thank you for these research topic ideas. It would (mostly) be researching at home, but related to markets I work on.
You had some very cool ideas in the past. Commodity markets have so many interesting domain details, I can't wait to hear what your next project will be!
Yes, commodities are very interesting markets! And merci for your kind words.
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

Re: Machine Learning/Neural Networks - Thijs Van Den Berg And Paul Wilmott - 30 Nov 2017 - London

March 30th, 2018, 8:56 am

Congratulations for the new ML course that will take place in London at the end of July. The course content is very appealing!

Course – Machine Learning With Thijs Van Den Berg And Paul Wilmott – London – 27-28 June 2018


Provisional Course Contents:
  • Unsupervised Learning models: Finding clusters[img=200x0]https://www.wilmott.com/wp-content/uplo ... 2017-1.jpg[/img]
  • Training a bot to play Blackjack
  • Speeding up pricing models with Neural Network function approximation
  • Modelling probability densities with Kernel Methods and Gaussian Mixture Models
  • Realistic simulation of Open/High/Low/Close volume bars
  • Extracting Stochastic Differential Equations from time series data with Neural Network function approximation
  • Modeling the dynamics of complex time series that have memory and non-Gaussian returns: How to outperform GARCH with Recurrent Neural Networks
  • Tools and techniques to prevent models blowing up
  • Non-linear dimension reduction with Autoencoders as a substitute for PCA
  • Building multi-factor Monte Carlo simulations that capture non-linear interactions with Variational Autoencoders
  • Simple and accurate arbitrage-free Implied Volatility smile modelling with GMM[img=200x0]https://www.wilmott.com/wp-content/uplo ... 2017-3.jpg[/img]
  • Modelling intraday volume and volatility with Recurrent Neural Networks
  • Pricing complex contracts with Reinforcement Learning methods
  • Optimizing investment strategies, learning the Kelly criterion from real-world data
  • Pairs trading, scanning portfolios for investment opportunities
  • Filling in missing data with denoising autoencoders