Reasoning data-driven

Chonghun Han, Ramachandran Lakshmanan, Bhavik Bakshi, and George Stephanopoulos, Nonmonotonic Reasoning The Synthesis of Operating Procedures in Chemical Plants Pedro M. Saraiva, Inductive and Analogical Learning Data-Driven Improvement of Process Operations... 

These techniques may be coarsely classified into techniques that mimic human reasoning about toxicological phenomena (Expert Systems) and methods that derive predictions from a training set of experimentally determined data (Data Driven Systems). 

Microwave-assisted synthesis is attractive to researchers for many reasons, including speed, yields, and the potential for reduced solvent use. Raman monitoring offers a convenient way to elucidate the chemical mechanism while instantly, continuously monitoring reaction kinetics. This enables rapid, data-driven process optimizations without concerns about safely and accurately sampling out of a microwave vessel stopped mid-reaction. Pivonka and Empheld of AstraZeneca Pharmaceuticals describe the continuous acquisition of Raman spectra of an amine or Knoevenagel coupling reaction in a sealed microwave reaction vessel at elevated temperatures and pressures [134]. 

Operating limits are data driven, with the most critical measure being the process variation. They are established by working inward from the registration limits to provide assurance that we will not fail registration limits because of process or assay noise. At this stage of development the process variation will not be completely known, particularly in the hands of the plant, but the noise observed from the DOEs can be a reasonable starting point. 

O Callaghan et al. [140]. Their analysis of seven CHO producer cell lines based on data-driven, in-depth protein formation models revealed that sufficient light chain (and also heavy chain) mRNA was produced but downstream protein formation was hampered for individual reasons in the cells. 

Ease of data acquisition Whether the system is rule-driven or data-driven , some data has to be acquired, even if this is just to help the rule-writer determine appropriate values for the rules. Here linear prediction clearly wins, because its parameters can easily be determined from any real speech w aveform. When formant synthesisers were mainly being developed, no fully reliable formant trackers existed, so the formant values had to be determined either manually or semi-manually. While better formant traekers now exist, many other parameters required in formant S5mthesis (e.g. zero loeations or bandwidth values) are still somewhat difficult to determine. Articulatory synthesis is partieularly interesting in that in the past it was next to impossible to acquire data. Now, various techniques such as EMA and MRI have made this much easier, and so it should be possible to collect much bigger databases for this purpose. The inability to collect accurate articulatory data is certainly one of the main reasons why articulatory synthesis never really took off. 

Text analysis A sensible starting point for any prediction would be to extrapolate current trends. To this extent, it is reasonable to assume that TTS will become entirely data driven. I think it is incontestable that the front end, or text processing component will become entirely statistical. In recent years the advances in statistical NLP have been enormous, fueled by the use of search engines and the need to translate documents. I think many of these techniques are directly applicable to TTS and will be adopted. 

Given the potential problems associated with both approaches, a third, middle way, has also been considered. This approach is called grey-box modelling, where the initial form of the equation determined based on the first-principle model is used for data-driven modelling. This approach has the advantage that the form of the equation has some physical meaning and could provide a reasonable description of the process. 

A few boards are tested, and, if results appear reasonable, the program is considered valid. The key drawback of the method is that self-learned test programs determine that all the boards are the same, not that they are good. Moreover, for economic fabrication of the test fixture, it is necessary to process product data anyway thus we may as well have output a data-driven test program. With virtually 100 percent of board designs being CAD-driven, there is little motivation to use self-learning today. 

---