Histopathology image classification can provide automated support towards cancer diagnosis. In this paper, we present a transfer learning-based approach for histopathology image classification. We first represent the image feature by Fisher Vector (FV) encoding of local features that are extracted using the Convolutional Neural Network (CNN) model pretrained on ImageNet. Next, to better transfer the pretrained model to the histopathology image dataset, we design a new adaptation layer to further transform the FV descriptors for higher discriminative power and classification accuracy. We used the publicly available BreaKHis image dataset for classifying between benign and malignant breast tumors, and obtained improved performance over the state-of-the-art.

We studied the physiological responses to abscisic acid (ABA) when 2-year-old potted plants of kiwifruit (Actinidia deliciosa) were grown under moisture stress. Leaves treated with 60 lM exogenous ABA through various means had less severe damage when water was limiting, and sprayed plants showed relatively greater drought resistance. This indicates that ABA improves tolerance in kiwifruit, reducing membrane permeability and enhancing the activities of antioxidant enzymes, e.g., peroxidase (POD), catalase (CAT), superoxide dismutase (SOD), ascorbate peroxidase (APX), and glutathione reductase (GR). Exposure to ABA led to higher levels of antioxidants, such as ABA and glutathione (GSH), while altering the amounts of endogenous hormones-ABA, indole-3-acetic acid (IAA), and Gibberellin (GA)-and organic oxalate, malate, and citrate in the leaves. Although daily applications of ABA were more effective than a single spray event, the effect of treatment, i.e., avoiding tissue damage and increasing plant resistance, was more apparent on Day 4 than on Day 6. No difference in response was apparent between control plants (regular irrigation) and those sprayed with ABA on Day 4 of the drought period.

In this paper, we investigate transmission of electromagnetic wave through aperiodic dielectric multilayers. A generic feature shown is that the mirror symmetry in the system can induce the resonant transmission, which originates from the positional correlations ͑for example, presence of dimers͒ in the system. Furthermore, the resonant transmission can be manipulated at a specific wavelength by tuning aperiodic structures with internal symmetry. The theoretical results are experimentally proved in the optical observation of aperiodic SiO 2 /TiO 2 multilayers with internal symmetry. We expect that this feature may have potential applications in optoelectric devices such as the wavelength division multiplexing system.

Fiber Bragg gratings (FBGs) are key elements in modern telecommunication and sensing applications. In optical communication, with the advancement of the Erbium doped fiber amplifier (EDFA), there is a great demand for devices with wavelength tunability over the Erbium gain bandwidth (in particular, for wavelength division multiplexing (WDM) networks). The center wavelength of a FBG can be shifted by means of change of temperature, pressure or mechanical axial strain. The axial strain approach is the best method among all other techniques because it allows relatively large wavelength shifts with high speed. Axial strain of up to 4% will be required to cover the whole EDFA region (more than 40 nm of central wavelength shift). The formation of Bragg grating results in significant reduction in mechanical strength of optical fibers especially in tension. As a result, axial strain of only about 1% can be achieved by mechanical stretching of FBGs. In order to achieve the remaining 3% strain compression of FBGs has to be applied.In this thesis, the design and analysis of a novel device for achieving central wavelength shift are presented. In particular, the device has achieved, for a fiber with Further, using the piezoelectric transducer (PZT) actuator as a driver, tuning speed of around 1.5nm/ms was achieved.-iii - ACKNOWLEDGMENTSUpon the completion of this thesis, I would like to express my sincere gratitude and appreciation to my supervisors Professor Chris W.

This paper concerns the development of a class of devices that generate end‐effector motion in the range of less than 100 μm and with sub‐nanometer resolution; in particular, a parallel manipulator configuration that generates a planar x‐y‐γ motion is considered. The parallel manipulator is implemented as a compliant mechanism. A problem with parallel manipulators is that the forward kinematics is usually too complex, which can hinder the implementation of advanced control algorithms. The contribution of this paper is that a simple method, called the constant‐Jacobian (CJ) method, is developed based on the pseudo‐rigid body (PRB) approach to compliant mechanisms. The experiment validates the CJ method. © 2002 John Wiley & Sons, Inc.