The presence of microorganisms, such as Escherichia coli, Salmonella, Listeria, and Vibrio, in food can be a serious threat to health, especially for infants, the elderly, and immunodeficient patients. The most reliable and accurate method for detecting food-borne pathogens is the conventional culture method, which includes a culture process and phenotypic or metabolic fingerprinting. [1] However, this method has the drawbacks that it is labor-intensive and requires at least 1-2 days to identify the pathogen(s). Therefore, researchers are looking for new methods that are fast, inexpensive, lightweight, and highly sensitive. Nanotechnology combined with biotechnology could potentially form the basis for such a method, and there have been efforts to develop fast and ultrasensitive nanosensors that can detect pathogens. To date, various platforms for pathogen detection have been developed, including metallic striped nanowires, [2] fluorescent nanobarcodes, [3] nanoparticles, [4,5] nanoelectromechanical systems (NEMs), [6] and microfluidic modules. [7,8] Herein, we demonstrate a screening tool for microorganisms such as E. coli, based on aptamer-functionalized single-walled carbon-nanotube field-effect transistor (SWNT-FET) arrays combined with the most probable number (MPN) method. Nanoscale biosensors based on FETs have been shown to be sufficiently sensitive to detect single viruses, [9] tumor-specific antigens, [10,11] and small molecules. [12] Moreover, in a previous study we showed that aptamer-functionalized SWNT-FET sensors can be used as sensitive, recyclable biosensors. [13] Nanosensors require only very small sample volumes (on the order of microliters), a characteristic that is advantageous in many instances but which can be a problem for inhomogeneous samples. Specifically, the probability that a collection of microorganisms in water will be distributed perfectly uniformly throughout the solution is very low. For example, if a 1-mL solution contains 10 3 E. coli cells, it does not necessarily mean that every 1-mL aliquot of the solution contains a single cell; rather, some aliquots will contain more than one cell, and others will not contain any cells. Thus, although each aliquot contains a single E. coli cell on average, there is a high possibility of recording a false signal if only a small volume of the solution is sampled. Microfluidic channels combined with nanosensors can solve this problem to some extent, but the volumes used are too small to avoid statistical errors. Moreover, motile bacteria such as E. coli can move at % 20 mm s À1 in a favorable medium, [14] which increases the difficulty of detecting these microorganisms using sensors with nanometer-sized sensing areas.Microbiologists have solved this problem by using a simple method called MPN. [15] First developed in 1933, MPN is still considered to be an important technique in estimating microbial populations in soils, waters, the food industry, etc. In MPN, to determine the cell titer in a particular solution, that solution is diluted at l...

No abstract

No abstract

Deep learning has achieved remarkable performance in various tasks thanks to massive labeled datasets. However, there are often cases where labeling large amount of data is challenging or infeasible due to high labeling cost such as labeling by experts or long labeling time per large-scale data sample (e.g., video, very large image). Active learning is one of the ways to query the most informative samples to be annotated among massive unlabeled pool. Two promising directions for active learning that have been recently explored are data distribution-based approach to select data points that are far from current labeled pool and model uncertainty-based approach that relies on the perspective of task model. Unfortunately, the former does not exploit structures from tasks and the latter does not seem to well-utilize overall data distribution. Here, we propose the methods that simultaneously take advantage of both data distribution and model uncertainty approaches. Our proposed methods exploit variational adversarial active learning (VAAL), that considered data distribution of both label and unlabeled pools, by incorporating learning loss prediction module and RankCGAN concept into VAAL by modeling loss prediction as a ranker. We demonstrate that our proposed methods outperform recent state-of-theart active learning methods on various balanced and imbalanced benchmark datasets.

No abstract