Deep learning for time series pyBCN
0 likes101 views
This document outlines a presentation on deep learning for time series analysis. It introduces the speaker and provides an outline of topics to be covered, including time series in various domains, classical time series analysis approaches, deep learning techniques like RNNs, CNNs and hybrid models, and takeaways. Classical approaches involve time domain, frequency domain and nearest neighbor algorithms as well as ARMA, smoothing and nonlinear dynamic models. Deep learning techniques discussed are RNNs, CNNs, combinations of RNNs and CNNs, and autoregressive CNNs. The presentation concludes with notes on hybrid solutions and takeaways around feature engineering, deep learning surpassing signal processing, and model combinations.
Deep learning for time series pyBCN
- 1. DEEP LEARNING for time series Alex Honchar | Mawi Solutions | Solutions architect, consultant
- 2. • 5 years in ML area | Ukraine, Russia, Italy, USA • AI Solution architect | Mawi Solutions, ECG analysis • AI consultant | self employed, finance and gaming ABOUT ME
- 3. • Time series in the wild • Classical approaches • Deep learning • Takeaways OUTLINE
- 4. TS IN THE WILD 🦁 • Planet Earth
- 5. TS IN THE WILD 🦁 • Planet Earth • Civilization delights
- 6. TS IN THE WILD 🦁 • Planet Earth • Civilization delights • Human beings
- 7. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis https://github.com/blue-yonder/tsfresh
- 8. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis https://github.com/blue-yonder/tsfresh
- 9. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis - nearest neighbors https://pypi.org/project/fastdtw/
- 10. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis - nearest neighbors - ML
- 11. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis - nearest neighbors - ML • Regression - ARIMA models http://www.statsmodels.org/
- 12. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis - nearest neighbors - ML • Regression - ARMA models - smoothing / decomposition https://pypi.org/project/stldecompose/
- 13. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis - nearest neighbors - ML • Regression - ARMA models - smoothing / decomposition - nonlinear dynamics
- 14. CLASSICAL APPROACHES 🔬 • Classification - time domain analysis - frequency domain analysis - nearest neighbors - ML • Regression - ARMA models - smoothing / decomposition - nonlinear dynamics - ML
- 15. DEEP LEARNING 🧠 TCE conference, 2014
- 16. DEEP LEARNING 🧠 • RNN
- 17. DEEP LEARNING 🧠 • RNN 1.Theoretical infinite memory 2.Multistep prediction ability 3.Don't work in parallel 4.Difficult to optimize 5.Slow in inference 6.Truncated implementation 7.Doubtful superior performance
- 18. DEEP LEARNING 🧠 • RNN • CNN
- 19. DEEP LEARNING 🧠 • RNN • CNN • RNN + CNN
- 20. DEEP LEARNING 🧠 • RNN • CNN • RNN + CNN
- 21. DEEP LEARNING 🧠 • RNN • CNN • RNN + CNN • Autoregressive CNN
- 22. DEEP LEARNING 🧠 • RNN • CNN • RNN + CNN • Autoregressive CNN • Other tasks
- 23. DEEP LEARNING 🧠 • RNN • CNN • RNN + CNN • Autoregressive CNN • Other tasks
- 25. TAKEAWAYS 📚 • There are dozens of features to feed classic ML with • Deep learning is eating signal processing • Autoregressive CNN > CNN > RNN • AEs and GANs are useful as well • Try to combine things!
- 26. Home reading 1. When Recurrent Models Don't Need To Be Recurrent 2. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling 3. Deep residual learning for image recognition 4. WaveNet: A generative model for raw audio 5. DEEP TEMPORAL CLUSTERING: FULLY UNSUPERVISED LEARNING OF TIME-DOMAIN FEATURES 6. REAL-VALUED (MEDICAL) TIME SERIES GENERATION WITH RECURRENT CONDITIONAL GANS 7. Time-series Extreme Event Forecasting with Neural Networks at Uber FB: @rachnogstyle IG: @rachnogstyle MEDIUM: @alexrachnog