About Us

Committed to speeding innovation for science-driven organizations, from product ideation through commercialization.


European Workshop on High-Throughput Developments & Applications

Accelrys Abstract

Talk - Developments in Data Management and Mining

Presented by: Dr. Dave Nicolaides, Principal Scientist, Accelrys
Authors: Dr Dave Nicolaides, Dr. Julian Willmott

Monday, February 27th, 5:50 pm
Track: Industry

The benefits of managing and mining the large amounts of data generated during HTE are well-known, yet the practice continues to be more difficult and less rewarding than it could be.  Here we present some hard truths which have emerged from the use of Accelrys software for these purposes by a selection of our customers over the recent years.  The truths include:

  • The necessity of making data management and mining future-proof.  This means supporting the inevitable situation that the information about what makes a good material often only emerges late in the development process, after you have performed many experiments which don’t provide that information.  „Just-in-time“ descriptor calculation is a method which can resolve this issue; we show how this works with examples, but do not omit to discuss its additional costs in terms of data storage.
  • The practice of data mining professionals doesn’t transfer easily to materials developers, who are simply not used to it.  Attempts to „embed expertise in software“ are ambitious undertakings, and must be accompanied by efforts to test and control the resulting analyses.  We present examples of several approaches here, some more successful than others.
  • HT and Low-throughput experimentation have to be integrated when doing data management and mining.  The current isolation of HT laboratories from the rest of the enterprise is really only another bottleneck in the R&D process.  We present examples where eliminating this bottleneck has provided significant value to our customers.

We also present recent internal results on efforts to unify design and analysis of HT experiments with design and analysis of computer experiments.  While both HTE and computer experiments share the goal of reducing the cost of experimentation, the details of these two approaches differ in ways which will require significant effort to resolve.

Back to Conference Page

Browse By: