I Explain Neural Networks to Life Insurance Actuaries

I was in Las Vegas recently for the Visual Studio Live conference. While I was there, I was invited to give a short informal talk at an American Council of Life Insurers (ACLI) event called Refocus 2020. The Refocus event had different tracks such as legal, compliance, and so on. I spoke to the actuarial sub-track.

I’ve been interested in actuarial science for many years because it applies mathematical and statistical methods to assess risk in insurance, finance and other industries and professions. But in some sense actuarial science is relatively primitive because it really boils down to statistical correlation in most cases.

One of the slides I used in my mini-presentation.

I spoke at a small out-of-band lunch time session. I used my standard presentation on neural networks, but because the members of the audience had very strong math backgrounds, I was able to dive deeper into the math details than I usually can. For example, I gave a fairly thorough explanation of tanh activation, softmax activation, data normalization, data encoding, and the Universal Approximation Theorem.

Based on my impromptu conversations with attendees, it’s my impression that actuarial scientists don’t use machine learning techniques much, if at all, but there’s tremendous interest in ML.

Three interesting old airline posters for Las Vegas.

This entry was posted in Conferences. Bookmark the permalink.