This work was supported by the BK-21 four program through the National Research Foundation of Korea (NRF) under the Ministry of Education. We would also like to present bundle of thanks to Nvidia Corporation for providing a support by donating us a Telsa K-40 GPU. Zubair khan and Farman Ali contributed equally and co-first authors ABSTRACT Diabetic retinopathy (DR) is a disease that damages retinal blood vessels and leads to blindness. Usually, colored fundus shots are used to diagnose this irreversible disease. The manual analysis (by clinicians) of the mentioned images is monotonous and error-prone. Hence, various computer vision hands-on engineering techniques are applied to predict the occurrences of the DR and its stages automatically. However, these methods are computationally expensive and lack to extract highly nonlinear features and, hence, fail to classify DR's different stages effectively. This paper focuses on classifying the DR's different stages with the lowest possible learnable parameters to speed up the training and model convergence. The VGG16, spatial pyramid pooling layer (SPP) and network-in-network (NiN) are stacked to make a highly nonlinear scale-invariant deep model called the VGG-NiN model. The proposed VGG-NiN model can process a DR image at any scale due to the SPP layer's virtue. Moreover, the stacking of NiN adds extra nonlinearity to the model and tends to better classification. The experimental results show that the proposed model performs better in terms of accuracy, computational resource utilization compared to state-of-the-art methods.

Sensor networks are handicapped by limited resources in the form of energy, processing, and memory. This paper proposes a new multi-hop energy efficient protocol, namely a routing algorithm using the ring-zone (RARZ) model. The protocol is lightweight, takes routing decisions based on the remaining energy of nodes, and performs location-based routing without the need for the nodes to know their respective positions. The protocol partitions the network into concentric rings around the base station. Each node assigns itself to a particular ring, known by a ringID. Multi-hop routing is performed and nodes within inner rings carry data for the outer rings towards the base station. Simulation results show that RARZ outperforms the address-light integrated MAC routing protocol (AIMRP), ad hoc on-demand distance vector (AODV) and Flooding in terms of end-to-end delay, average hop count, and energy consumption.

Abstract:In order to lower the dependence on textual annotations for image searches, the content based image retrieval (CBIR) has become a popular topic in computer vision. A wide range of CBIR applications consider classification techniques, such as artificial neural networks (ANN), support vector machines (SVM), etc. to understand the query image content to retrieve relevant output. However, in multi-class search environments, the retrieval results are far from optimal due to overlapping semantics amongst subjects of various classes. The classification through multiple classifiers generate better results, but as the number of negative examples increases due to highly correlated semantic classes, classification bias occurs towards the negative class, hence, the combination of the classifiers become even more unstable particularly in one-against-all classification scenarios. In order to resolve this issue, a genetic algorithm (GA) based classifier comity learning (GCCL) method is presented in this paper to generate stable classifiers by combining ANN with SVMs through asymmetric and symmetric bagging. The proposed approach resolves the classification disagreement amongst different classifiers and also resolves the class imbalance problem in CBIR. Once the stable classifiers are generated, the query image is presented to the trained model to understand the underlying semantic content of the query image for association with the precise semantic class. Afterwards, the feature similarity is computed within the obtained class to generate the semantic response of the system. The experiments reveal that the proposed method outperforms various state-of-the-art methods and significantly improves the image retrieval performance.

No abstract

Multivariate chemical reaction optimization involving catalytic systems is a non-trivial task due to the high number of tuneable parameters and discrete choices. Closed-loop optimization featuring active Machine Learning (ML) represents a powerful strategy for automating reaction optimization. However, the translation of chemical reaction conditions into a machine-readable format comes with the challenge of finding highly informative features which accurately capture the factors for reaction success and allow the model to learn efficiently. Herein, we compare the efficacy of different calculated chemical descriptors for a high throughput generated dataset to determine the impact on a supervised ML model when predicting reaction yield. Then, the effect of featurization and size of the initial dataset within a closed-loop reaction optimization was examined. Finally, the balance between descriptor complexity and dataset size was considered. Ultimately, tailored descriptors did not outperform simple generic representations, however, a larger initial dataset accelerated reaction optimization.