O slideshow foi denunciado.
Utilizamos seu perfil e dados de atividades no LinkedIn para personalizar e exibir anúncios mais relevantes. Altere suas preferências de anúncios quando desejar.

The Move to AI: from HFT to Laplace Demon

212 visualizações

Publicada em

Video and slides synchronized, mp3 and slide download available at URL http://bit.ly/2o2Ffr2.

Eric Horesnyi and Albert Bifet discuss how hedge funds have moved beyond High Frequency Trading using AI and real-time data processing. Filmed at qconlondon.com.

Eric Horesnyi is the CEO of streamdata.io. He was a founding team member at Internet Way, then Radianz (Global Finance Cloud, sold to BT). He is a High Frequency Trading infrastructure expert, passionate about Fintech, IoT and Cleantech. Albert Bifet is Associate Professor at Telecom ParisTech and Honorary Research Associate at the WEKA Machine Learning Group at University of Waikato.

Publicada em: Tecnologia
  • Seja o primeiro a comentar

  • Seja a primeira pessoa a gostar disto

The Move to AI: from HFT to Laplace Demon

  1. 1. From HFT to Laplace Demon When timed data technology curves the market @erichoresnyi @abifet
  2. 2. InfoQ.com: News & Community Site Watch the video with slide synchronization on InfoQ.com! https://www.infoq.com/presentations/ ai-high-frequency-trade • Over 1,000,000 software developers, architects and CTOs read the site world- wide every month • 250,000 senior developers subscribe to our weekly newsletter • Published in 4 languages (English, Chinese, Japanese and Brazilian Portuguese) • Post content from our QCon conferences • 2 dedicated podcast channels: The InfoQ Podcast, with a focus on Architecture and The Engineering Culture Podcast, with a focus on building • 96 deep dives on innovative topics packed as downloadable emags and minibooks • Over 40 new content items per week
  3. 3. Purpose of QCon - to empower software development by facilitating the spread of knowledge and innovation Strategy - practitioner-driven conference designed for YOU: influencers of change and innovation in your teams - speakers and topics driving the evolution and innovation - connecting and catalyzing the influencers and innovators Highlights - attended by more than 12,000 delegates since 2007 - held in 9 cities worldwide Presented at QCon London www.qconlondon.com
  4. 4. HFT in the hey days 5ms=20m$Source: Tabb Group High Frequency Trading
  5. 5. $100trn CapGroup Vanguard StateStreet BoNYJPMPimco Fidelity GS Prudential AUM>$1trn, source: Towers Watson Blackrock* *Blackrock is actually headquartered in NY, main AUM coming from ETF/ passive originally BGI in SF Approx 3xGDP in USA ie 155k$/hab HFT context
  6. 6. Liquidity Flow Buy Side 2 Buy Side 1 NASDAQ NYSENSX Sell Side 2 Sell Side 1 70% Algo PCX HFT context
  7. 7. Order Flow NY NJ CTIL Sell Side 2Sell Side 1 Market Maker CA HFT context
  8. 8. PCX INET BRUT NASDAQ CME CBOT ICEBATS NYSE ARCA CBOE NYMEX IEX Cambrian Explosion Reg.ATS'98-Reg.NMS'05 HFT context
  9. 9. Infra view NJ NJ NY IL CO POP Fiber 20ms 2ms 2ms 4ms HFT context
  10. 10. 1 HFT: Proximity NY NJ CT IL 16 2 1 1 Host in Network Nodes, then Exchanges HFT
  11. 11. + Serialization Latency = + Processing Propagation
  12. 12. HFT: Ultra NY NJ CT IL HFT Dark Fiber 1115
  13. 13. Buy-Side view of HFT
  14. 14. It's not a ghost...
  15. 15. HFT: Straight Fiber 1,000 miles > 825 miles 14.5 ms > 11.5 ms
  16. 16. HFT: Microwaves 11.5 > 8.5ms N:1.33 > 1.0003 v = c/n
  17. 17. HFT: FPGANanosecs
  18. 18. Choose your lane HFT <> Algo Trading "Once you get into milliseconds it's almost not HFT any more"
  19. 19. Spacetime is relative Market Events: [ct,x,y,z]
  20. 20. Speed curves spacetime HFT built a wormhole to win on [ct',x,y,z] events
  21. 21. Mass curves spacetime AI builds a blackhole by massively processing [ct,x,y,z] events G + Λg = Tµν µν c4 8πG µν {
  22. 22. Laplace Demon The endgame of Determinism ∀ [ct,x,y,z] ∈ Rn ⊢ ∀ [ct',x',y',z']
  23. 23. The Endgame 1/3 Event Machine View
  24. 24. The Endgame 2/3 Loss aka Cost Function = J(θ) : distance points to line Graph View : Regression
  25. 25. ⎣ ⎢ ⎢ ⎡ w w ...w1,1 1,2 1,n w w ...w2,1 2,2 2,n w w ...w...,1 ...,2 ...,n w w ...wn,1 n,2 n,n ⎦ ⎥ ⎥ ⎤ The Endgame 3/3 [x , x , ..., x0 1 n] Features Labels $AAPL $GOOG . $FB ⎣ ⎢ ⎢ ⎡y1 y2 . yn ⎦ ⎥ ⎥ ⎤ Matrix view Matrices of Weights
  26. 26. AI not news to trading +35% yoy for 20 years : $2,500 > $1,000,000 PhD Mathematics, Berkeley - String Theory Chern-Simons Form
  27. 27. AI age:Gradient Descent Follow the steepest slope, 100m+ features α : Learning Rate, ∇ J : Gradient
  28. 28. AI age:Back Propagation Adapt weight to control error from previous layer's input, 150+ layers Source: Neural Networks simulation by Matt Mazur at Emergent Mind
  29. 29. AI age: GPU From Final Fantasy to Autonomous Car "The implementation of streaming algorithms, typied by highly parallel computations with little reuse of input data, has been widely explored on GPUs."(Stanford, 2004)
  30. 30. Bullish Fitness Drill 3-Test 2-Validate 1-Train Overfitting?
  31. 31. Bearish Fitness Drill 3-Test 2-Validate 1-Train Overfitting?
  32. 32. Standard Approach Batch-based, finite training sets, static models Dataset Model
  33. 33. Data Stream Approach Infinite training sets, dynamic models D D D D D D MM M M M M
  34. 34. Approximation Algo What is the largest number that we can store in 8 bits?
  35. 35. Approximation Algo What is the largest number that we can store in 8 bits?
  36. 36. Approximation Algorithm
  37. 37. Massive Online Analysis
  38. 38. Stream Setting Process an example at a time Inspect it only once (at most) Use a limited amount of memory Work in a limited amount of time Be ready to predict at any point
  39. 39. Prequential Evaluation Sequence of examples > Error of a model
  40. 40. Command Line java -cp .:moa.jar:weka.jar -javaagent:sizeofag.jar moa.DoTask EvaluatePrequential -l DecisionStump //training DecisionStump classifier ... -s generators.WaveformGenerator //...on WaveformGenerator data -n 100000 //using the first 100 thousand examples for testing -i 100000000 //training on a total of 100 million examples -f 1000000 //testing every one million examples > dsresult.csv
  41. 41. Resourceful Classification Regression Concept Drift Sentiment Analysis Stock Price Alerting
  42. 42. Simple learner.getVotesForInstance(instance) learner.trainOnInstance(instance)
  43. 43. Scalablehttp://samoa-project.net
  44. 44. An experiment Public Stock Dataset MOA Regression Stock Price Error
  45. 45. An Experiment
  46. 46. The New HF Frontier: AI {API} Sentiment Analysis Alerts Regression/Perceptron
  47. 47. Fast vs Smart Data Stream a compromise ct x,y,z HFT AIData Stream
  48. 48. Thanks! Apache & Wikipedia Foundation : please donate! MOA, Kaggle & Giphy : please contribute! Books & Lectures Data Stream Mining, MOA team Yann LeCun Deep Learning Class, NYU Matt Mazure, Emergent Mind & Andrew Ng, Coursera on AI My Life as a Quant:Reflections on Physics&Finance, E.Derman The Value of a Millisecond: Finding the Optimal Speed of a Trading Infra., TabbGroup Flashboys, M.Lewis Movies & Games The Big Short, Back to the Future, Interstellar, The Black Hole, Harry Potter, Rocky, Into the Mind, Star Wars, Matrix; Final Fantasy
  49. 49. Watch the video with slide synchronization on InfoQ.com! https://www.infoq.com/presentations/ai-high- frequency-trade

×