Articles

Data science at the core of the digital revolution

With the advent of Big Data, data science per se has become a research field in its own right, where corporate investments can be seen to be quite ‘massive’ too. Dr Benjamin Quost, research scientist and lecturer at UTC-Heudiasyc Lab, updates Interactions’ readers on the latest trends observed in this area.

Data science at the core of the digital revolution

Applications based on analysis of recurring features in data bases with millions of examples and hundreds of thousands of variables - as found, for instance, in industrial service sector, in energy … offer some hitherto unknown opportunities. “Statisticians used to complain they didn’t have enough data, whereas today the new challenge often relates to the sheer quantity of information to be processed”, notes Benjamin Quost. Starting with the first academic publications on the topic in the early 2000s, the number of research projects and associate investments have literally taken off. Following a period of experimentation, over the past 4 to 5 years, the technologies involved have matured and now pervade a great many areas of activities. Creation of powerful, efficient algorithms and ever-increasing processing power levels and speeds have multiplied tenfold the capacity for scientists to analyse situations and have even in certain to solve the problems without human intervention. “Previous concepts, such as neuron networks have been perfected; we have now progressed to self-learning machines, for example by combining deep learning process and reinforcement Pavlov learning”, underscores Benjamin Quost, who reminds us of the success of the Alphago game programme, winning against the world’s finest Go champions, something considered to unthinkable 15 years ago. Among other applications that use vast amounts of data provided by sensors and users are the ‘smart’, driverless cars developed by colleagues at UTC-Heudiasyc also rely, to a large extent, on new possibilities to be found in artificial intelligence (AI). The level of expertise attained by this joint UTC-CNRS research laboratory, created in 1981, designates it as a key player in research which often receives assignments and contract offers from its entrepreneurial partners. This unit brings with it a definite strategic added value.

Applications in day-to-day life

Over and above “High Tech” applications, current solutions that can readily be accessed via personal computers (PCs), connected support devices or on-board sensors are already available on sale or being developed. As lecturer Quost stresses, “Enterprises propose projects as varied as connected buildings or posture detection via sensor equipped tee-shirts”. Inasmuch as Dr Quost is in charge of the elective specialty “data mining and decision-making”, he has been able to observe at first hand the level of enthusiasm displayed by students in regard to Big Data as a topic. What we see as a trend in data science to replace human operators by machines could also lead to a future social upheaval. Moreover, the spectacular progress noted in data science relate to fields where the quantities of data available are stupendous. Some of the general problems are characterized by limited data inputs. “Research and associate investments are focused a lot on Big Data, but we should not neglect Small data, where we endeavour to analyse incomplete data or data with noise.” he explains. Making models on the basis of “poor” information sources represents a promising step forward in areas such as biology and medicine, where the data can indeed be limited or incomplete, often with uncertain quality and this can be important when it comes to making decisions.