Stochastic Gradient Descent (SGD) is the central workhorse for training modern CNNs. Although giving impressive empirical performance it can be slow to converge. In this paper we explore a novel strategy for training a CNN using an alternation strategy that offers substantial speedups during training. We make the following contributions: (i) replace the ReLU non-linearity within a CNN with posi...

In this paper, a deep neural network (Behavior-CNN) is proposed to model pedestrian behaviors in crowded scenes, which has many applications in surveillance. A pedestrian behavior encoding scheme is designed to provide a general representation of walking paths, which can be used as the input and output of CNN. The proposed Behavior-CNN is trained with real-scene crowd data and then thoroughly i...

Metadata Title: CNN News Update Description: The latest news happening in the U.S. and around the world. Episode 1 Title: CNN News Update (8-21-2007 7 AM EDT) MP3: http://rss.cnn.com/...08-21-07-7AM.mp3 Episode 2 Title: CNN News Update (8-21-2007 6 AM EDT) MP3: http://rss.cnn.com/...08-21-07-6AM.mp3 Episode 3 Title: CNN News Update (8-21-2007 5 AM EDT) MP3: http://rss.cnn.com/...08-21-07-5AM.mp...

A robust testing method for detecting circuit faults within two-dimensional Cellular Neural Network (CNN) arrays is presented. The functional tests consist of a sequence of input vectors that toggle all internal nodes of the conceptual CNN model and propagate the result to the output pins. The resultant output vectors reveal nodes that exhibit opened, shorted, or stuck-at faults. The generated ...

Convolutional neural networks (CNN) currently dominate the computer vision landscape. Recently, a CNN based model, Faster R-CNN [1], achieved stateof-the-art performance at object detection on the PASCAL VOC 2007 and 2012 datasets. It combines region proposal generation with object detection on a single frame in less than 200ms. We apply the Faster R-CNN model to video clips from the ImageNet 2...

ABSTRACT We report on the design of a new full-analog current-mode CNN in a 1.2 pm CMOS technology, whose cell core is characterized by an intrinsic capability of weights control, low power consumption and small area occupation. Circuit simulations allowed the design approach to be validated and the electrical performance of the CNN to be predicted; moreover, it is shown that the proposed CNN c...

Convolutional neural networks (CNN) have achieved the top performance for event detection due to their capacity to induce the underlying structures of the k-grams in the sentences. However, the current CNN-based event detectors only model the consecutive k-grams and ignore the non-consecutive kgrams that might involve important structures for event detection. In this work, we propose to improve...

Purpose Real-time assessment of ventricular volumes requires high acceleration factors. Residual convolutional neural networks (CNN) have shown potential for removing artifacts caused by data undersampling. In this study we investigated the effect of different radial sampling patterns on the accuracy of a CNN. We also acquired actual real-time undersampled radial data in patients with congenita...

Early diagnosis is playing an important role in preventing progress of the Alzheimer's disease (AD). This paper proposes to improve the prediction of AD with a deep 3D Convolutional Neural Network (3D-CNN), which can show generic features capturing AD biomarkers extracted from brain images, adapt to different domain datasets, and accurately classify subjects with improved fine-tuning method. Th...

In the field of medicine, with the introduction of computer systems that can collect and analyze massive amounts of data, many non-invasive diagnostic methods are being developed for a variety of conditions. In this study, our aim is to develop a non-invasive method of classifying respiratory sounds that are recorded by an electronic stethoscope and the audio recording software that uses variou...