Error Measurement and "Cost of Forecast Error" for Predictive Models

Hello,

I'm currently working on multivariate regression using Artificial Neural Networks to forecast the price of VIX. My problem is that I'm an undergraduate Math student and not a trader so I don't know which cost function I should be minimizing.

I usually default to RMSE for regression problems, but I understand there are a number of other metrics used in statistics for error measurement of forecasts.
(eg.
Mean square error (MSE): Sum of the squared errors
Root mean square error (RMSE): Square root of the MSE
Mean percent error (MPE): Average of the percent errors
Mean absolute error (MAE): Average of the absolute errors
Mean absolute percent error (MAPE): Average of the APE
Weighted mean absolute percent error (WMAPE): Weighted average of the APE)
source: goo.gl/nyYaAy

I will eventually be summarizing my findings in a paper and hope to submit the results to the SIAM Journal on Financial Mathematics (SIFIN) for publication.

I found a paper which published a number of different error metrics in a nice table (ie. goo.gl/3ZWpst) but in terms of training the neural network, the optimizer needs a single objective to minimize.

As a trader, if you were going to include a forecasting model such as an ANN in your trading strategy, which error measurement would you be most interested in seeing minimized? Would this answer change if you were trading FOREX vs. Options vs. Stocks?

Thank you for any guidance you can provide!