A5029500893- Neural Networks and Applications
- Neural dynamics and brain function
- EEG and Brain-Computer Interfaces
- Functional Brain Connectivity Studies
- Underwater Acoustics Research
- Face and Expression Recognition
- Visual perception and processing mechanisms
- Blind Source Separation Techniques
- Speech and Audio Processing
- Advanced Vision and Imaging
- Image Retrieval and Classification Techniques
- Target Tracking and Data Fusion in Sensor Networks
- Phonocardiography and Auscultation Techniques
- Face Recognition and Perception
- ECG Monitoring and Analysis
- Structural Health Monitoring Techniques
- Image Processing and 3D Reconstruction
- Image and Signal Denoising Methods
- Neural and Behavioral Psychology Studies
- Fault Detection and Control Systems
- Bat Biology and Ecology Studies
- Marine animal studies overview
- Visual Attention and Saliency Detection
- Advanced Memory and Neural Computing
- Handwritten Text Recognition Techniques
Tel Aviv University
2014-2023
Geriatric Research Education and Clinical Center
2021
Brown University
1999-2010
Allen Institute for Brain Science
2010
University of Bologna
2005
John Brown University
2002-2005
Gustavus Adolphus College
2004
Exact Sciences (United States)
1995
University of Pennsylvania
1987
Abstract Relating brain tissue properties to diffusion tensor imaging (DTI) is limited when an image voxel contains partial volume of with free water, such as cerebrospinal fluid or edema, rendering the DTI indices no longer useful for describing underlying properties. We propose here a method separating from surrounding water while mapping volume. This achieved by fitting bi‐tensor model which mathematical framework introduced stabilize fitting. Applying on datasets healthy subject and...
Views Icon Article contents Figures & tables Video Audio Supplementary Data Peer Review Share Twitter Facebook Reddit LinkedIn Tools Reprints and Permissions Cite Search Site Citation Mohamad H. Hassoun, Nathan Intrator, Susan McKay, Wolfgang Christian; Fundamentals of Artificial Neural Networks. Comput. Phys. 1 March 1996; 10 (2): 137. https://doi.org/10.1063/1.4822376 Download citation file: Ris (Zotero) Reference Manager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex toolbar...
Based on an observation about the different effect of ensemble averaging bias and variance portions prediction error, we discuss training methodologies for ensembles networks. We demonstrate reduction present a method extrapolation to limit infinite ensemble. A significant is obtained by just over initial conditions neural networks, without varying architectures or sets. The minimum error reached later than that single network. In vicinity minimum, appears be flatter network, thus...
Bootstrap samples with noise are shown to be an effective smoothness and capacity control technique for training feedforward networks other statistical methods such as generalized additive models. It is that noisy bootstrap performs best in conjunction weight-decay regularization ensemble averaging. The two-spiral problem, a highly non-linear, noise-free data, used demonstrate these findings. combination of averaging also useful modelling, demonstrated on the well-known Cleveland heart data.
Electroencephalographic (EEG) analysis has emerged as a powerful tool for brain state interpretation and diagnosis, but not the diagnosis of mental disorders; this may be explained by its low spatial resolution or depth sensitivity. This paper concerns schizophrenia using EEG, which currently suffers from several cardinal problems: it heavily depends on assumptions, conditions prior knowledge regarding patient. Additionally, diagnostic experiments take hours, accuracy is unreliable. article...
We examine a classification problem in which seismic waveforms of natural earthquakes are to be distinguished from man-made explosions. present an integrated machine (ICM), is hierarchy artificial neural networks (ANNs) that trained classify the waveforms. In order maximize gain combining multiple ANNs, we suggest construction redundant environment (RCE) consists several "experts" whose expertise depends on different input representations they exposed. proposed scheme, experts ensembles ANN,...
This work presents a novel method for automatic detection and identification of heart sounds. Homomorphic filtering is used to obtain smooth envelogram the phono cardiogram, which enables robust events interest in sound signal. Sequences features extracted from detected are as observations hidden Markov model. It demonstrated that task major sounds can be learned unlabelled cardiograms by an unsupervised training process without assistance any additional synchronizing channels
A novel unsupervised neural network for dimensionality reduction that seeks directions emphasizing multimodality is presented, and its connection to exploratory projection pursuit methods discussed. This leads a new statistical insight into the synaptic modification equations governing learning in Bienenstock, Cooper, Munro (BCM) neurons (1982). The importance of principle based solely on distinguishing features demonstrated using phoneme recognition experiment. extracted are compared with...
We present a new supervised learning procedure for ensemble machines, in which outputs of predictors, trained on different distributions, are combined by dynamic classifier combination model. This may be viewed as either version mixture experts (Jacobs, Jordan, Nowlan, & Hinton, 1991), applied to classification, or variant the boosting algorithm (Schapire, 1990). As experts, it can made appropriate general classification and regression problems initializing partition data set boostlike...
There is interest in extending the boosting algorithm (Schapire, 1990) to fit a wide range of regression problems. The threshold-based for used an analogy between classification errors and big regression. We focus on practical aspects this compare it other attempts extend capabilities model are demonstrated laser data from Santa Fe times-series competition Mackey-Glass time series, where results surpass those standard ensemble average.
Abstract The problem of representing the spatial structure images, which arises in visual object processing, is commonly described using terminology borrowed from propositional theories cognition, notably, concept compositionality. classical stance mandates representations composed symbols, stand for atomic or composite entities and enter into arbitrarily nested relationships. We argue that main desiderata a representational system—productivity systematicity—can (indeed, number reasons,...
One of the greatest challenges involved in studying brain mechanisms fear is capturing individual's unique instantaneous experience. Brain imaging studies to date commonly sacrifice valuable information regarding individual real-time conscious experience, especially when focusing on elucidating amygdala's activity. Here, we assumed that by using a minimally intrusive cue along with applying robust clustering approach probe amygdala, it would be possible rate real time and derive related...
It is known that acoustic heart sounds carry significant information about the mechanical activity of heart. In this paper, we present a novel type cardiac monitoring based on sound analysis. Specifically, study two morphological features and their associations with physiological changes from baseline state. The framework demonstrated recordings during laparoscopic surgeries 15 patients. Insufflation, which performed surgery, provides controlled, externally induced stress, enabling an...