Market Consistency and WMC [27]

Go to: Summary | Previous | Next   
Bullet points include: The lambda_j and hence the p_i can then be found using standard optimisation techniques. E.g. conjugate gradient, DFP, BFGS, L-BFGS, L-BFGS-B minimisation / optimisation algorithms, see e.g. Press et al. (2007). Algorithms that utilise the (multi-dimensional) gradient of the relative entropy function (plus boundary penalties) are appealing, given computational effort required to calculate the gradients versus an individual function evaluation. The problem can also be vectorised and hence parallelised, if suitable parallelisation software and e.g. access to GPUs are available. Again, similarities with machine learning and related techniques. A good route for actuarial students to get ahead of the pack along the ‘data science’ curve?

Contents | Prev | Next | Library

Desktop view | Switch to Mobile