Should Quants Learn More About Machine Learning?

We want systems that understand financial markets and anticipate their behavior. Quantitative finance is about the development of models and computational methods and their applications.

A Workout in Computational Finance emphasizes on numerical schemes as thorough grounding to everything. Because if the calculations go wrong everything goes wrong - often financial equations need to be solved many million of times to get a principle picture of the behavior.

Riding the price waves we take the price dynamic as absolute and invert into something. A model, hopefully. We say we calibrate it by determining its parameters such that its forecasts fit well to the prices we observed.

The explanation of the remaining differences between the real price waves and that forecast by the  calibrated model we could call context.

One aspect of machine learning is about the extraction of context from data. An understandable model. If it is also computational it can be used to close the gaps automatically. Simplified, if this then apply that model instance.

Again, if your calibrations go wrong this contextual technology goes wrong.

What is the difference of doing a machine learning project from a quant development project?

In both you deal with data and data analysis as an exploratory approach should be the begin of everything. But this is not what I mean here.

In machine learning you may create models from data (if you know nothing), you may create models that close the delta between the reality and calculations (if you have a (nonparametric) model), or you create a model that adapts your model parameters.

In quant finance we have models and they are even calibrated to market data. In riding the price waves we will do recalibration constantly to novel price structures.

To build an adaptive technology (understandable and computational models) would need (much) more influencing information than our models for the movements of our underlyings have.

But how to get this extra data, things like profit and debt graphs, ... to do a kind of contextual computational finance?

What will be the contextual technologies that will help us?

Big financial data? Technologies like Google's PageRank algorithms? New computing muscles? Mobile devices?

Machine learning will help ... But applying its adaptive technologies will need a sound foundation -  accurate, fast and robust mathematical schemes for valuations ... that might need to be applied billions of times to get one better decision.

Picture from sehfelder