Deep Learning is undergoing intense study as it continues to attain outstanding results because of multiple levels of abstractions. Currently, deep learning algorithms are trained and executed on a single machine with multiple Graphics Processing Units (GPUs). To achieve better results, a large amount of data need to be trained in deep neural network with expensive GPUs and General Purpose GPU (GPGPUs). Training complex deep learning models take weeks and sometimes even months. In order to converge faster without comprising in accuracy percentage, distributed and parallel nature of Apache Spark comes into play. Spark has an advantage of in-memory and fast processing of data. Recent researchers involve the integration of deep learning and Apache Spark to exploit computation power and scalability. In this paper, all the recent deep learning frameworks are reviewed exhaustively with a detailed assessment and comparison. Experimentation results are provided to evaluate which frameworks are more suitable for deep models. We also discussed the open issues related to each deep learning framework.
In the pharmaceutical field, early detection of lung nodules is indispensable for increasing patient survival. We can enhance the quality of the medical images by intensifying the radiation dose. High radiation dose provokes cancer, which forces experts to use limited radiation. Using abrupt radiation generates noise in CT scans. We propose an optimal Convolutional Neural Network model in which Gaussian noise is removed for better classification and increased training accuracy. Experimental demonstration on the LUNA16 dataset of size 160 GB shows that our proposed method exhibit superior results. Classification accuracy, specificity, sensitivity, Precision, Recall, F1 measurement, and area under the ROC curve (AUC) of the model performance are taken as evaluation metrics. We conducted a performance comparison of our proposed model on numerous platforms, like Apache Spark, GPU, and CPU, to depreciate the training time without compromising the accuracy percentage. Our results show that Apache Spark, integrated with a deep learning framework, is suitable for parallel training computation with high accuracy.
22 February 2019 in International Journal of Interactive Mobile Technologies (iJIM)
Touch interface technologies for mobile devices are essentially in use. The purpose of such touch interfaces is to run an application by touching a screen with a user’s finger or to implement various functions on the device. When the user has an attempt to use the touch interface, users tend to grab the mobile device with one hand. Because of the existence of untouchable areas to which the user cannot reach with the user’s fingers, it is possible to occur for a case where the user is not able to touch a specific area on the screen accurately. This results in some issues that the mobile device does not carry out the user’s desired function and the execution time is delayed due to the wrong implementation. Therefore, there is a need to distinguish the area where the user can stably input the touch interface from the area where the users cannot and to overcome the problems of the unstable touch area. Furthermore, when the size of the screen increases, these issues will become more serious because of an increase in the unstable touch areas. Especially, an interface that receives position and force data like 3D-touch requires the stable area setting different from the conventional 2D-touch. In this paper, we search and analyze the stable touch areas on the large screen where the user can do 3D-touch inputs.