RAM’s

Machine learning

A new dimension
to financial data

Machine Learning

Artificial intelligence (“AI”) has arguably become the most significant and disruptive general purpose technology in recent years.

Techniques like machine learning (“ML”) have taken giant strides forward; empowering computers and enabling them to learn and build models so that they can perform prediction across different domains.

This technology is particularly pertinent to the portfolio management industry, where the integration of information from numerous, large – and often unstructured – datasets has become a key success factor.

ML development frameworks accessible to the research community now offer solutions to both handle these datasets and interpret results, controlling how inferences are derived. We believe that what we teach the machine is only as important as the limits and constraints we set during its learning process, unlocking interpretability and generalization.

In this paper we will look at the growing popularity of artificial intelligence and the potential for finance to prosper from machine learning implementations. As technology continues to push the boundaries of our imagination, new dimensions will undoubtedly emerge over time.

Machine Learning

What’s driving the growing
popularity of machine learning?

This current AI boom is rooted in research from the 1950’s

1950’s

Artificial
Intelligence

From a
concept ...

1980’s

Machine
Learning

To a set of
predictive techniques

2000’s

Deep
Learning

Learning in always more levels of abstraction

1950
1951
1965
1982
1997
2015 to…

The phenomenal acceleration of ML techniques is certainly no fad

Web-search data from Google Trends provides an interesting insight into what’s driving the recent enthusiasm for machine learning techniques:

Google Trends - Interest Study

Machine Learning

What is
machine learning?

Our Approach

Why do we use
machine learning?

Our Approach

How do we control
the machine?

Conclusion

The general enthusiasm for machine learning techniques will not turn out to be a short-lived fashion trend. It is proving to be, across numerous fields of application, the most efficient tool to extract information and build predictive models from large datasets.

The significant investments of tech giants into the technology and hardware supporting deep learning will help maintain this trend. The flexibility and robustness of their open-source development frameworks, as well as the significant processing power accessible (either directly or through their highly-evolutive cloud infrastructures), make machine learning techniques efficient data processing techniques. This is increasingly welcome at a time when digital data is exploding, alternative or unstructured financial datasets.
But we must be conscious that the history of artificial intelligence is awash with over-exuberant expectations.

Here, the most stunning example could be the optimistic forecast of cognitive scientist Minsky in 1970 : “In from three to eight years we will have a machine with the general intelligence of an average human being”. Who would only accurately predict today when fully autonomous self-driving cars will be reliable ?

We prefer to regard machine learning techniques as a generalization of traditional data processing techniques, and our research efforts are equally focused on testing models and controlling them by making sure they generalize and provide tangible results. At a time when the number of attempts to generate truly human-like AI is increasing and excitation about fully-autonomous vehicles is growing, we embrace initiatives about “Human trust in AI”.

RAM-AI

Tel. +41 58 726 87 00

E-mail contact@ram-ai.com

Contact Us