Stochastic Newton-type optimization, unlike deterministic Newton optimisation, uses noisy (possibly approximate) 2nd-order gradient information to find the argument which minimises

\[ x^*=\operatorname{argmin}_{\mathbf{x}} f(x) \]

for some an objective function \(f:\mathbb{R}^n\to\mathbb{R}\).

## Subsampling

Most of the good tricks here are set up for ML-style training losses where the bottleneck is summing a large number of loss functions.

LiSSA attempts to make 2nd order gradient descent methods scale to large parameter sets (Agarwal, Bullins, and Hazan 2016):

a linear time stochastic second order algorithm that achieves linear convergence for typical problems in machine learning while still maintaining run-times theoretically comparable to state-of-the-art first order algorithms. This relies heavily on the special structure of the optimization problem that allows our unbiased hessian estimator to be implemented efficiently, using only vector-vector products.

David McAllester observes:

Since \(H^{t+1}y^t\) can be computed efficiently whenever we can run backpropagation, the conditions under which the LiSSA algorithm can be run are actually much more general than the paper suggests. Backpropagation can be run on essentially any natural loss function.

(Kovalev, Mishchenko, and Richtárik 2019) uses a decomposition of the objective into a sum of simple functions (the classic SGD setup for neural nets, typical of an online optimisation.

What does Francis Bach’s finite sample guarantee research get us in this setting? (F. Bach and Moulines 2011; F. R. Bach and Moulines 2013)

Many machine learning and signal processing problems are traditionally cast as convex optimization problems. A common difficulty in solving these problems is the size of the data, where there are many observations ('large n') and each of these is large ('large p'). In this setting, online algorithms such as stochastic gradient descent which pass over the data only once, are usually preferred over batch algorithms, which require multiple passes over the data. In this talk, I will show how the smoothness of loss functions may be used to design novel algorithms with improved behavior, both in theory and practice: in the ideal infinite-data setting, an efficient novel Newton-based stochastic approximation algorithm leads to a convergence rate of O(1/n) without strong convexity assumptions, while in the practical finite-data setting, an appropriate combination of batch and online algorithms leads to unexpected behaviors, such as a linear convergence rate for strongly convex problems, with an iteration cost similar to stochastic gradient descent. (joint work with Nicolas Le Roux, Eric Moulines and Mark Schmidt).

## General case

Rather than observing \(\nabla f, \nabla^2 f\) we observe some random variables \(G(x),H(x)\) with \(bb{E}G=\nabla f\) and \(\bb{E}(H)=\nabla^2 f,\) not necessarily decomposable into a sum.

🏗

For now, see optimisation of experiments which is an interesting application of this problem.

## References

*Advances in Neural Information Processing Systems (NIPS)*, –. Spain. http://hal.archives-ouvertes.fr/hal-00608041.

*Neural Computation*4 (2): 141–66. https://doi.org/10.1162/neco.1992.4.2.141.

*Journal of Machine Learning Research*10 (December): 1737–54. http://jmlr.org/papers/volume10/bordes09a/bordes09a.pdf.

*Neural Networks: Tricks of the Trade*, 421–36. Lecture Notes in Computer Science. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-35289-8_25.

*SIAM Journal on Optimization*26 (2): 1008–31. https://doi.org/10.1137/140954362.

*Advances In Neural Information Processing Systems*. http://arxiv.org/abs/1509.03475.

*Advances in Neural Information Processing Systems 27*, 2933–41. Curran Associates, Inc. http://arxiv.org/abs/1406.2572.

*Proceedings of the 27th International Conference on International Conference on Machine Learning*, 735–42. ICML’10. USA: Omnipress. http://www.cs.utoronto.ca/~jmartens/docs/Deep_HessianFree.pdf.

*Proceedings of the 28th International Conference on International Conference on Machine Learning*, 1033–40. ICML’11. USA: Omnipress. http://dl.acm.org/citation.cfm?id=3104482.3104612.

*Neural Networks: Tricks of the Trade*, 479–535. Lecture Notes in Computer Science. Springer. http://www.cs.toronto.edu/~jmartens/docs/HF_book_chapter.pdf.

*Optimizing Methods in Statistics*, edited by Jagdish S. Rustagi, 233–57. Academic Press. https://doi.org/10.1016/B978-0-12-604550-5.50015-8.

*The Annals of Statistics*13 (1): 236–45. https://doi.org/10.1214/aos/1176346589.

*Artificial Intelligence and Statistics*, 436–43. http://proceedings.mlr.press/v2/schraudolph07a.html.