# Syllabus DQE
### Irina Espejo

## **1.** **Gaussian Processes**
- Introduction
- Properties of different kernels and flexibility
- Scalability and fast posterior fit
[7] Wang et al. 2019
[10] Gardner et al. 2018

---

## **2.** **Bayesian Optimization**
- Black Box optimization
[6] J. Snoek et al. 2012
[8] Hernandez-Lobato et al. 2014 (PES)
<!-- [9] Zanette et al. 2018 -->
- Multiarmed bandits and Reinforcement Learning (huge, cut down)
- Decision theory
- Back to LFI how to connect it all, connection with excursion project

---

## **3.** **Likelihood Free Inference**
- Why the likelihood can be intractable
- Traditional methods and its curses:
- ABC method
[1] Rubin DB, 1984 and
[2] Beaumont et al., 2002.
<!-- Extensive review in [3] Sisson SA, 2018.-->
- Histograms and kernel-based estimation models
- Advanced simulation-based inference trends [4] Cranmer et al. 2020
- Normalizing flows, autoregressive models
- Active learning
- Integration and augmentation
- Application to particle physics: MadMiner [11], [12]
- Pros and cons of the techniques above (curse of dimensionality, amortization..)

---

## **4.** **The role of workflows**
[5] Chen et al. 2019
[12] Cranmer et al. 2011
[13] Cranmer et al. 2017
[14] Cranmer et al. 2018

- Compartmentalization: often complex simulation-based inference pipelines are an ensemble of simpler steps
- Keep the scientist out of the pipeline as much as possible
- Black box functions can be wrapped as workflows and called during the LFI pipeline multiple times
- Simulators (black boxes) may have complex dependencies
- Reusability and amortization
- REANA
- Different use cases of workflows in science and AI:
The National Academies of Science Egineering and Medicine, "Realizing Opportunities for Advanced and Automated Workflows in Scientific Research: Second Meeting"

- Biomedical research: Robert F. Murhpy, [slides](https://www.nationalacademies.org/event/03-16-2020/docs/D616442DE6CA39C1515CDF881E04DC8EA8DDD9F4583C)
- Material Science: Tyrel McQueen [slides](https://www.nationalacademies.org/event/03-16-2020/docs/DEC311C5C4401B82455B5884F158CF6F8A821FF1EA1D)
- Open science: Beth Plale [slides](https://www.nationalacademies.org/event/03-16-2020/docs/D3795B07818DF9019C3B0AE2E478BAF2E1988DCDD110)
- Acceleration and automation of science: Ian Foster [slides](https://www.nationalacademies.org/event/03-16-2020/realizing-opportunities-for-advanced-and-automated-workflows-in-scientific-research-second-meeting#sectionEventMaterials)

---

## References

[1] Rubin, Donald B. “Bayesianly Justifiable and Relevant Frequency Calculations for the Applied Statistician.” _The Annals of Statistics_ 12, no. 4 (December 1984): 1151–72 [DOI](https://doi.org/10.1214/aos/1176346785)

[2] M. A. Beaumont, W. Zhang, and D. J. Balding: "Approximate bayesian computation in population genetics". _Genetics_ 162 (4), p. 2025, 2002 [link](https://www.genetics.org/content/162/4/2025)

<!--[3] Sisson SA (2018) Handbook of Approximate Bayesian Computation. (Chapman andHall/CRC).-->

[4] Cranmer, K., Brehmer, J., & Louppe, G. (2020). The frontier of simulation-based inference. _Proceedings of the National Academy of Sciences_ [DOI](https://doi.org/10.1073/pnas.1912789117)

[5] Snoek, J., Larochelle, H., & Adams, R. (2012). Practical Bayesian Optimization of Machine Learning Algorithms. In _Proceedings of the 25th International Conference on Neural Information Processing Systems - Volume 2_ (pp. 2951–2959). Curran Associates Inc. [DOI](https://doi.org/10.1038/s41567-018-0342-2)

[6] J. Snoek, H. Larochelle, and R. P. Adams. Practical bayesian optimization of machine learning algorithms. In _Advances in neural information processing systems_, pages 2951–2959, 2012

[7] K.A. Wang, G. Pleiss, J.R.Gardner, S.Tyree, K.Q.Weinberg, A.G.Wilson. Exact Gaussian Processes on a Million Data Points. In _Advances in neural information processing systems_, 2019.

[8] J.M.Hernandez-Lobato, M.W.Hoffman, Z.Ghahramani. Predictive entropy search for efficient global optimization of black-box functions. In _Advances in neural information processing systems_, pages 918-926, 2014.

<!--[9] A.Zanette, J.Zhang, M.Kochenderfer. Robust Super-Level Set Estimation using Gaussian Processes. ECML,2018. In _Advances in neural information processing systems_, 2018.-->

[10] J.R.Gardner, G.Pleiss, D.Bendel, K.Weinberger, A.G.Wilson. GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration. In _Advances in neural information processing systems_, pages 7587–7597, 2018.

[10] J.Brehmer, K.Cranmer, G.Louppe, J.Pavez. Constraining Effective Field Theories with Machine Learning. _Physical Review Letters_ 121, (2018).

[11] J.Brehmer, F.Kling, I.Espejo, K.Cranmer. MadMiner: Machine learning-based inference for particle physics. _Computing and Sotware for Big Science 4_ (2020).

[12] Cranmer, K. & Yavin, I. RECAST — extending the impact of existing analyses. _J. High Energy Phys._ 2011, 38 (2011)

[13] Cranmer, Kyle, and Itay Yavin. “RECAST — Extending the Impact of Existing Analyses.” _Journal of High Energy Physics_ 2011, no. 4 (April 2011): 38. [DOI](https://doi.org/10.1007/JHEP04)(2011)038.

[14] Cranmer, Kyle, Lukas Heinrich, and ATLAS collaboration. “Analysis Preservation and Systematic Reinterpretation within the ATLAS Experiment.” _Journal of Physics: Conference Series_ 1085 (September 2018): 042011. [DOI](https://doi.org/10.1088/1742-6596/1085/4/042011).

---