*TBD.*

I'm working through a small realisation, for my own interest, which has been helpful in my understanding of variational Bayes; specifically relating it to non-Bayes variational inference. Also sequential monte carlo.

By starting from the idea of sufficient statistics, we come to the idea of variational inference in a natural way, via some other interesting stopovers.

I doubt this insight is novel, but I will work through it as if it is, for the sake of my own education.

I will contend that most Bayesian inference (maybe *also* frequentist, but
the setup here is easier with Bayes) is most naturally considered as filtering.
We have some system who parameters we wish to estimate, and we have many
experiments done upon that system, throughout time. We would like to update our
understanding of that system, using the experiments' data. Possibly the system
is changing, possibly it's not changing. Either way (or both ways) there is a
computational challenge.

(Or it's not changing, but our model is increasing in size as we gain more data about it. How to handle this case?)

We would like the hypothesis updating to happen in a compact way, where we can do inference without keeping the whole data-set about, but rather use some summary statistics. With exponential family models this is trivial to do; just use conjugate updates and you are done. With models where compactness would be more useful, say, big data and big models without conjugate updates, it's not clear how we can do this exactly.

See also mixture models, probabilistic deep learning, directed graphical models, other probability metrics. Intuitive connection with differential privacy of posterior sampling? DNZM13

Possibly related:
Data summarisation
via *inducing sets* or *coresets*.
Bounded Memory Learning?
To mention: Bayesian likelihood principle,
Pitman–Koopman–Darmois theorem.

## Sufficient statistics in exponential families

Let's start with sufficient statistics in exponential families, which, for reasons of historical pedagogy, are the Garden of Eden of Inference, the Garden of Edenference for short. I suspect that deep in their hearts, all statisticians regard themselves as prodigal exiles from of the exponential family, and long for the innocence of that Garden of Edenference.

Anyway, informally speaking, here's what's going on with the inference problems involving sufficient statistics. We are interested in estimating some parameter of inference, using realisations of some random process

Then is a *sufficient statistic* for iff
That is, our inference about depends on the data *only* through the
sufficient statistic.

(mention size of sufficient statistic)

Fisher–Neyman factorization theorem

Famously, Maximum Likelihood estimators of
exponential family models
are highly compressible, in that these have *sufficient statistics* -
these are low-dimensional functions of the data which characterise
all the information in the complete data,
with respect to the parameter estimates.
Many models and data sets and estimation methods do not have this feature,
even parametric models with very few parameters.

This can be a PITA when your data is very big and you wish to get benefit from that, and yet you can't fit the data in memory; The question then arises – when can I do better? Can I find a “nearly sufficient” statistic, which is smaller than my data and yet does not worsen my error substantially? Can I quantify this nearness to the original?

## Refs

- AdCo09: Adragni, K. P., & Cook, R. D. (2009) Sufficient dimension reduction and prediction in regression.
*Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences*, 367(1906), 4385–4405. DOI - BaLK15: Bachem, O., Lucic, M., & Krause, A. (2015) Coresets for Nonparametric Estimation – the Case of DP-Means. In International Conference on Machine Learning (pp. 209–217).
- Baez11: Baez, J. C. (2011) Renyi Entropy and Free Energy.
- BaRY98: Barron, A., Rissanen, J., & Yu, B. (1998) The minimum description length principle in coding and modeling.
*IEEE Transactions on Information Theory*, 44(6), 2743–2760. DOI - BaSW99: Barron, Andrew, Schervisch, M. J., & Wasserman, L. (1999) The Consistency of Posterior Distributions in Nonparametric Problems.
*The Annals of Statistics*, 27(2), 536–561. DOI - BlKM17: Blei, D. M., Kucukelbir, A., & McAuliffe, J. D. (2017) Variational Inference: A Review for Statisticians.
*Journal of the American Statistical Association*, 112(518), 859–877. DOI - CKMY16: Cortes, C., Kuznetsov, V., Mohri, M., & Yang, S. (2016) Structured Prediction Theory Based on Factor Graph Complexity. In D. D. Lee, M. Sugiyama, U. V. Luxburg, I. Guyon, & R. Garnett (Eds.), Advances in Neural Information Processing Systems 29 (pp. 2514–2522). Curran Associates, Inc.
- CBMF17: Cutajar, K., Bonilla, E. V., Michiardi, P., & Filippone, M. (2017) Random Feature Expansions for Deep Gaussian Processes. In PMLR.
- DeBo15: Dezfouli, A., & Bonilla, E. V. (2015) Scalable Inference for Gaussian Process Models with Black-box Likelihoods. In Advances in Neural Information Processing Systems 28 (pp. 1414–1422). Cambridge, MA, USA: MIT Press
- DNZM13: Dimitrakakis, C., Nelson, B., Zhang, and Z., Mitrokotsa, A., & Rubinstein, B. (2013) Bayesian Differential Privacy through Posterior Sampling.
*ArXiv:1306.1066 [Cs, Stat]*. - GBKT17: Gribonval, R., Blanchard, G., Keriven, N., & Traonmilin, Y. (2017) Compressive Statistical Learning with Random Feature Moments.
*ArXiv:1706.07180 [Cs, Math, Stat]*. - Grün04: Grünwald, P. D. (2004) A tutorial introduction to the minimum description length principle.
*Advances in Minimum Description Length: Theory and Applications*, 23–81. - HiCa93: Hinton, G. E., & van Camp, D. (1993) Keeping the Neural Networks Simple by Minimizing the Description Length of the Weights. In Proceedings of the Sixth Annual Conference on Computational Learning Theory (pp. 5–13). New York, NY, USA: ACM DOI
- HoVa04: Honkela, A., & Valpola, H. (2004) Variational learning and bits-back coding: an information-theoretic view to Bayesian learning.
*IEEE Transactions on Neural Networks*, 15(4), 800–810. DOI - KKPW14: Kandasamy, K., Krishnamurthy, A., Poczos, B., Wasserman, L., & Robins, J. M. (2014) Influence Functions for Machine Learning: Nonparametric Estimators for Entropies, Divergences and Mutual Informations.
*ArXiv:1411.4342 [Stat]*. - KuLe51: Kullback, S., & Leibler, R. A. (1951) On Information and Sufficiency.
*The Annals of Mathematical Statistics*, 22(1), 79–86. - Mack02: MacKay, D. J. C. (2002)
*Information Theory, Inference & Learning Algorithms*. Cambridge University Press - Mand62: Mandelbrot, B. (1962) The Role of Sufficiency and of Estimation in Thermodynamics.
*The Annals of Mathematical Statistics*, 33(3), 1021–1038. DOI - Mont15: Montanari, A. (2015) Computational implications of reducing data to sufficient statistics.
*Electronic Journal of Statistics*, 9(2), 2370–2390. DOI - MoMo17: Moshkovitz, D., & Moshkovitz, M. (2017) Mixing Implies Lower Bounds for Space Bounded Learning. In Conference on Learning Theory (pp. 1516–1566).
- MoTi17a: Moshkovitz, M., & Tishby, N. (2017a) A General Memory-Bounded Learning Algorithm.
*ArXiv:1712.03524 [Cs]*. - MoTi17b: Moshkovitz, M., & Tishby, N. (2017b) Mixing Complexity and its Applications to Neural Networks.
*ArXiv:1703.00729 [Cs]*. - Riss78: Rissanen, J. (1978) Modeling by shortest data description.
*Automatica*, 14(5), 465–471. DOI - Riss84: Rissanen, J. (1984) Universal coding, information, prediction, and estimation.
*IEEE Transactions on Information Theory*, 30(4), 629–636. DOI - Riss07: Rissanen, Jorma. (2007)
*Information and complexity in statistical modeling*. New York: Springer - Vitá06: Vitányi, P. M. (2006) Meaningful information.
*IEEE Transactions on Information Theory*, 52(10), 4617–4626. DOI - Wall90: Wallace, C. S. (1990) Classification by minimum-message-length inference. In Advances in Computing and Information — ICCI ’90 (pp. 72–81). Springer, Berlin, Heidelberg DOI
- Wolp08: Wolpert, D. H. (2008) Physical limits of inference.
*Physica D: Nonlinear Phenomena*, 237(9), 1257–1281. DOI