A self-adaptive synthetic over-sampling technique for imbalanced classification

Traditionally, in supervised machine learning, (a significant) part of the available data (usually 50%-80%) is used for training and the rest—for validation. In many problems, however, the data are highly imbalanced in regard to different classes or does not have good coverage of the feasible data space which, in turn, creates problems in validation and usage phase. In this paper, we propose a technique for synthesizing feasible and likely data to help balance the classes as well as to boost the performance in terms of confusion matrix as well as overall. The idea, in a nutshell, is to synthesize data samples in close vicinity to the actual data samples specifically for the less represented (minority) classes. This has also implications to the so-called fairness of machine learning. In this paper, we propose a specific method for synthesizing data in a way to balance the classes and boost the performance, especially of the minority classes. It is generic and can be applied to different base algorithms, for example, support vector machines, k-nearest neighbour classifiers deep neural, rule-based classifiers, decision trees, and so forth. The results demonstrated that (a) a significantly more balanced (and fair) classification results can be achieved and (b) that the overall performance as well as the performance per class measured by confusion matrix can be boosted. In addition, this approach can be very valuable for the cases when the number of actual available labelled data is small which itself is one of the problems of the contemporary machine learning.     

Proteomic Helicobacter pylori biomarkers discriminative of low-grade gastric MALT lymphoma and duodenal ulcer

To date no reliable diagnostic method exists to predict, among the very large and clinically heterogeneous group of Helicobacter pylori‐infected patients, the extremely small group at risk for developing low‐grade gastric MALT lymphoma (LG‐MALT). Search of proteomic biomarkers holds promise for the classification of the H. pylori strains with regard to this severe clinical outcome. In the present study 69 H. pylori strains isolated from patients with two different H. pylori‐associated diseases, duodenal ulcer (DU, n=29) and LG‐MALT (n=40) were used. Protein expression patterns of the strains were analyzed by using the high‐throughput methodology SELDI. Selected proteins were purified by means of chromatographic and electrophoretic methods in view of further sequencing by LC‐MS/MS. Univariate analysis (Mann–Whitney test) of the protein expression patterns generated nine significant biomarkers that can discriminate between H. pylori strains from patients with DU and LG‐MALT. These biomarkers are of low molecular weight, ranging from 6 to 26.6 kDa. Among them, two are overexpressed in LG‐MALT strains and seven – in DU strains. Two biomarker proteins, one overexpressed in LG‐MALT strains (13.2 kDa) and another one – overexpressed in DU strains (26.6 kDa), were purified to homogeneity and identified by using LC‐MS/MS as a 50S ribosomal protein L7/L12 and a urease subunit, respectively. These biomarkers can be included in novel protein arrays for the differential diagnosis of H. pylori‐associated clinical outcomes.     

Towards a Mobile Code Management Environment for Complex,Real-Time,Distributed Systems

We present a novel mobile code management environment,currently under design and development. Our design employs anopen architecture, suitable for ``plug-and-play' with COTS andother groups' tools. While we have studied new algorithms, costand objective functions, and other fundamental issues, the maincontribution of this experimental research work is in the environmentitself. It should be noted that networked platforms, such asthe World Wide Web, are inherently not suitable for traditional,predictable real-time applications. Thus, real-time concernsnecessarily need to be blended with others concerns, and thetarget applications, making use of our environment, will toobe a blend of partially hard real-time and partially (or mostly)soft-real- time ones. The prototype environment will thereforesupport performance-based analysis and management focusing notonly on predictability but also on compilation, efficiency, safetyand other tradeoffs. We have selected the Java language and itsbytecode format as a representation for mobile code as well asa language for our implementation.     

A rapid differentiation between oak honeydew honey and nectar and other honeydew honeys by NMR spectroscopy

An NMR-based approach for rapid differentiation of oak honeydew honey from all other honey types (floral and other honeydew honeys) was proposed. It is based on the identification of the signals of the protons and the carbon of the methylene group of quercitol in the ¹H and ¹³C NMR spectra of honey. The presence of quercitol was supported by TOCSY spectroscopy. Quercitol is a deoxyinositol which is regarded as a good taxonomic marker for the genus Quercus. All samples of oak honeydew honey contained quercitol, while in floral honey samples and honeydew honeys from fir and spruce it was absent. In addition, the described approach is promising with respect to quercitol quantification in honey by qNMR.     

Improving dissolution kinetics of pharmaceuticals by fluidized bed impregnation of active pharmaceutical ingredients

Investigational drugs are increasingly becoming less soluble in aqueous media, thus, presenting real challenges during development. Previous work has successfully demonstrated the manufacturing of pharmaceuticals using fluidized bed (FB) impregnation of APIs onto porous carriers. This study demonstrates the usefulness of FB impregnation in formulating poorly soluble drugs. We show that dissolution of Fenofibrate is greatly improved by FB impregnation onto Neusilin^® (Fuji Health Science Inc, Burlington, NJ USA), a synthetic amorphous form of magnesium alumino‐metasilicate. We impregnate Neusilin^® for range of loadings and examine Fenofibrate's physical state. Dissolution of impregnated formulations is drug loading dependent and loadings below 40% show great improvement (decrease) in release time compared to physical blend. Release times are further improved by milling. We also examine feasibility of coimpregnating Fenofibrate with additives and observe stability (1.5 years) of the amorphous form of Fenofibrate inside Neusilin^®. This stabilization significantly improves Fenofibrate's dissolution kinetics, making our formulation comparable to one of the current market formulations, TriCor^® tablets (AbbVie Inc, North Chicago, IL USA). © 2016 American Institute of Chemical Engineers AIChE J, 62: 4201–4214, 2016     

Fuzzy systems design: direct and indirect approaches

A systematic classification of the data-driven approaches for design of fuzzy systems is given in the paper. The possible ways to solve this modelling and identification problem are classified on the basis of the optimisation techniques used for this purpose. One algorithm for each of the two basic categories of design methods is presented and its advantages and disadvantages are discussed. Both types of algorithms are self-learning and do not require interaction during the process of fuzzy model design. They perform adaptation of both the fuzzy model structure (rule-base) and the parameters. The indirect approach exploits the dual nature of Takagi-Sugeno (TS) models and is based on recently introduced recursive clustering combined with Kalman filtering-based procedure for recursive estimation of the parameter of the local sub-models. Both algorithms result in finding compact and transparent fuzzy models. The direct approach solves the optimisation problem directly, while the indirect one decomposes the original problem into on-line clustering and recursive estimation problems and finds a sub-optimal solution in real-time. The later one is computationally very efficient and has a range of potential applications in real-time process control, moving images recognition, autonomous systems design etc. It is extended in this paper for the case of multi-input–multi-output (MIMO systems). Both approaches have been tested with real data from an engineering process.     

Performance of glucose biosensor based on oxygen electrode in physiological fluids and at body temperature

A potentially implantable glucose biosensor for continuous monitoring of glucose levels in diabetic patients has been developed. The glucose biosensor is based on an amperometric oxygen electrode and glucose oxidase immobilized on carbon powder held in the form of a liquid suspension. The enzyme material can be replaced (the sensor recharged) without sensor disassembly. Diffusion membranes made of silastic coatings over a microporous polycarbonate membrane are used. Calibration curves of the sensors in phosphate buffer solution and in undiluted blood plasma at body temperature have been obtained. The reproducibility of the sensor response in serum at body temperature is demonstrated. The sensors have a stable signal during storage and continuous operation at body temperature for a period of at least one month.     

Effects of nickel foam dimensions on catalytic activity of supported Co–Mn–B nanocomposites for hydrogen generation from stabilized borohydride solutions

In the present work, the catalytic activity of electrodeposited Co–Mn–B nanocomposites towards controllable hydrolysis of sodium borohydride was studied. Deposition was performed on two types of Ni-foam (RECEMAT Int.) with different pore size, specific surface area and thickness. Higher deposit loading, as well as bigger real surface area, was obtained with foam samples possessing bigger pore size. The catalyst deposited on bigger pore foam promoted hydrogen generation with higher rates than the other one when contacted with a base-stabilized NaBH₄ solution. The same activation energy value, however, was determined for both supported catalysts. On the base of the obtained results, it may be concluded that the geometric factor plays predominant role for the catalytic activity of studied catalysts.     

Eine Implementierung von Losgröße 1 nach Industrie-4.0-Prinzipien

The synergy of machine and IT and the resulting business opportunities and models represent key features of Industrie 4.0, a high prioritized governmental initiative focusing on further expanding the leading role of German machine and plant construction. The ongoing standardization efforts are a prerequisite for cross-vendor communication. Nevertheless, early practical experiments, realistic field trials and pilot projects in partnership play a great role in the rapprochement between the two worlds and in particular for the goal-oriented use of those innovations. To show the great importance of the software, which is seen as a driver of innovation in mechanical engineering, an example was built by an IBM team led by the two authors that demonstrates the batch size 1 scenario from Industrie 4.0 based on a 3D printer. A customer orders his individual product on a web shop, a production order is created, which—after approval—is directly sent to production. The machine data are displayed live in the context of the job on a production control console respectively in an augmented reality app as well as collected for the purpose of deeper insights and forecasts. Flexible vertical integration of machines, their integration into business processes across the entire value chain as well as the insights gained from the production data are some Industrie 4.0 scenarios, which can be illustrated very well in this example.