I was reading this news about DeepMind first commercial initiative and was thinking about what would be the impact of such Google cloud structure on stocks or even Forex market.
In my opinion, a great challenge to any AI product, and of course professionals and organizations, is not regarding just on supervised learning and very narrow application, like this relevant case.
And I hope at the near future see DeepMind similar experiences on trading, because here we really have the future infinity uncertainty and complexity for both, people and machines.
Or, in other words, the real challenge to any AI solution.
Here is the quote from Dr. Ernest P. Chan http://www.epchan.com/biography/ about whether machine learning models has been proved successful or not in quantitative trading:
I do not find machine learning methods to be generally useful in quantitative trading. (I have explained this view in my first book Quantitative Trading.) The reason is that most machine learning methods tend to have too many parameters that need to be optimized with in-sample data. But financial time series data is actually quite limited (going high frequency doesn't help much due to the highly correlated nature of such data). So the amount of in-sample data isn't enough to overcome data snooping bias, and most machine learning models fail out-of-sample in financial time series predictions. Also, financial market suffers frequent regime changes, so few machine learning models with complicated rules can survive those changes without understanding the deeper, more fundamental patterns than only humans can detect and utilize (due to their contextual/external knowledge).
Dr. Ernest P. Chan is the Managing Member of QTS Capital Management, LLC. His career since 1994 has been focusing on the development of statistical models and advanced computer algorithms to find patterns and trends in large quantities of data. He has applied his expertise in statistical pattern recognition to projects ranging from textual...
In the last post in our machine learning series, we showed how nonlinear regression algos might improve regression forecasting relative to plain vanilla linear regression (i.e., when underlying reality is nonlinear with complex interactions). In this piece, we’ll first review machine learning for classification, a problem which may be less...
In the new book Advances in Financial Machine Learning by Marcos Lopez De Prado he proposes that quant researchers utilize a different type of price bar.His research has shown that by using alternatives to fixed time interval bars (minute, hour, day, week, etc.), the return series will exhibit better statistical properties. In other words using alternative bar types, the return series will better approximate normality/stationarity which will make our research and conclusions more robust.
Introduction This post explores a concept at the heart of quantitative financial research. Most qfin researchers utilize statistical techniques that require varying degrees of stationarity. As many of you are aware financial time series violate pretty much all the rules of stationarity and yet many researchers, including me, have applied or...
Just a small list of links to discussion about risks and opportunities of robots in quantitative funds, and opportunities for developers in the field, especially for those who are still skeptical about the use of AI in the market.
Introduction “The question is not whether intelligent machines can have any emotions, but whether machines can be intelligent without any emotions”. Marvi…
Here is the quote from Dr. Ernest P. Chan http://www.epchan.com/biography/ about whether machine learning models has been proved successful or not in quantitative trading:
I do not find machine learning methods to be generally useful in quantitative trading. (I have explained this view in my first book Quantitative Trading.) The reason is that most machine learning methods tend to have too many parameters that need to be optimized with in-sample data. But financial time series data is actually quite limited (going high frequency doesn't help much due to the highly correlated nature of such data). So the amount of in-sample data isn't enough to overcome data snooping bias, and most machine learning models fail out-of-sample in financial time series predictions. Also, financial market suffers frequent regime changes, so few machine learning models with complicated rules can survive those changes without understanding the deeper, more fundamental patterns than only humans can detect and utilize (due to their contextual/external knowledge).
In the new book Advances in Financial Machine Learning by Marcos Lopez De Prado he proposes that quant researchers utilize a different type of price bar. His research has shown that by using alternatives to fixed time interval bars (minute, hour, day, week, etc.), the return series will exhibit better statistical properties. In other words using alternative bar types, the return series will better approximate normality/stationarity which will make our research and conclusions more robust.
http://www.blackarbs.com/blog/exploring-alternative-price-bars
https://towardsdatascience.com/financial-machine-learning-part-0-bars-745897d4e4ba
Just a small list of links to discussion about risks and opportunities of robots in quantitative funds, and opportunities for developers in the field, especially for those who are still skeptical about the use of AI in the market.
Billionaire Robots: Machine Learning at Renaissance Technologies
https://rctom.hbs.org/submission/billionaire-robots-machine-learning-at-renaissance-technologies/
“The Massive Hedge Fund Betting on AI”. Adam Satariano and Nishant Kumar
https://www.bloomberg.com/news/features/2017-09-27/the-massive-hedge-fund-betting-on-ai
“Renaissance Technologies Co-Founder Reveals Secrets to Quant and VC Career Success”. Dan Butcher
https://news.efinancialcareers.com/us-en/298218/renaissance-technologies-secrets-to-quant-hedge-funds-vc-career-success
The Revolutionary Way Of Using Artificial Intelligence In Hedge Funds
https://www.forbes.com/sites/bernardmarr/2019/02/15/the-revolutionary-way-of-using-artificial-intelligence-in-hedge-funds-the-case-of-aidyia/#5da632eb57ca
Million Dollar Salaries for AI Researchers? Well, we Quants Have Seen this Movie Before
https://blog.goodaudience.com/million-dollar-salaries-for-ai-researchers-well-we-quants-have-seen-this-movie-before-8e7af51f6c63
How hedge funds are using AI and machine learning
https://www.computerworlduk.com/galleries/data/how-hedge-funds-are-using-ai-machine-learning-3672971/
https://www.datasciencecentral.com/profiles/blogs/can-ai-detect-emotions-better-than-humans
Forum on trading, automated trading systems and testing trading strategies
Machine learning, neural networks, deep learning : Do you find the available material on this site adequate ?
Alain Verleyen, 2018.05.13 20:29