A5060342498- Neural Networks and Applications
- Machine Learning and Algorithms
- Natural Language Processing Techniques
- Digital Media Forensic Detection
- Domain Adaptation and Few-Shot Learning
- Topic Modeling
- Machine Learning and Data Classification
- Image and Signal Denoising Methods
- Handwritten Text Recognition Techniques
- Face and Expression Recognition
- Evolutionary Algorithms and Applications
- Bayesian Modeling and Causal Inference
- Data Quality and Management
- Image Processing Techniques and Applications
- Advanced Image Processing Techniques
- Sparse and Compressive Sensing Techniques
- Metaheuristic Optimization Algorithms Research
- Web Data Mining and Analysis
- Stochastic Gradient Optimization Techniques
- Video Analysis and Summarization
- Fractal and DNA sequence analysis
- Multimodal Machine Learning Applications
- Global Financial Crisis and Policies
- Advanced Neural Network Applications
- Machine Learning in Bioinformatics
Ernst & Young (United States)
2021-2023
Ernst & Young (Israel)
2020-2021
Art Institute of Portland
2021
Sentient Science (United States)
2017-2018
University of California, Santa Cruz
1999-2003
DNA microarray experiments generating thousands of gene expression measurements, are being used to gather information from tissue and cell samples regarding differences that will be useful in diagnosing disease. We have developed a new method analyse this kind data using support vector machines (SVMs). This analysis consists both classification the samples, an exploration for mis-labeled or questionable results.We demonstrate detail on consisting ovarian cancer tissues, normal other tissues....
This paper introduces new learning algorithms for natural language processing based on the perceptron algorithm. We show how can be efficiently applied to exponential sized representations of parse trees, such as "all subtrees" (DOP) representation described by (Bod 1998), or a tracking all sub-fragments tagged sentence. give experimental results showing significant improvements two tasks: parsing Wall Street Journal text, and named-entity extraction from web data.
Removing noise from scanned pages is a vital step before their submission to optical character recognition (OCR) system. Most available image denoising methods are supervised where the pairs of noisy/clean required. However, this assumption rarely met in real settings. Besides, there no single model that can remove various types documents. Here, we propose unified end-to-end unsupervised deep learning model, for first time, effectively multiple noise, including salt & pepper blurred and/or...
In many evolutionary optimization domains evaluations are noisy. The candidates tested on a number of randomly drawn samples, such as different games played, physical simulations, or user interactions. As result, selecting the winner is multiple hypothesis problem: candidate that evaluated best most likely received lucky selection and will not perform well in future. This paper proposes technique for estimating its true performance based smoothness assumption: Candidates similar similarly....
The first workshop on Document Intelligence (DI-2019) was held December 14, 2019 at NeurIPS conference in Vancouver, Canada. report summarizes the workshop, with a summary of talks, papers and posters presented, discusses common themes, issues open questions that came up workshop.
The success of deep learning depends on finding an architecture to fit the task. As has scaled up more challenging tasks, architectures have become difficult design by hand. This paper proposes automated method, CoDeepNEAT, for optimizing through evolution. By extending existing neuroevolution methods topology, components, and hyperparameters, this method achieves results comparable best human designs in standard benchmarks object recognition language modeling. It also supports building a...
We introduce DeepPSL a variant of probabilistic soft logic (PSL) to produce an end-to-end trainable system that integrates reasoning and perception. PSL represents first-order in terms convex graphical model – hinge-loss Markov random fields (HL-MRFs). stands out among frameworks due its tractability having been applied systems more than 1 billion ground rules. The key our approach is represent predicates using deep neural networks then approximately back-propagate through the HL-MRF thus...
In the regular course of business, companies spend a lot effort reading and interpreting documents, highly manual process that involves tedious tasks, such as identifying dates names or locating presence absence certain clauses in contract. Dealing with natural language is complex further complicated by fact these documents come various formats (scanned image, digital formats) have different degrees internal structure (spreadsheets, invoices, text documents). We present DICR, an end-to-end,...
Computer security has always been an issue, more so in recent years due to global network access. In this paper, we present a simple connectionist algorithm for testing the quality of computer passwords. A popular method evaluating password is test it against large dictionary words and near-words. Our approximate realization method. The stored distributed form. All are stable; however, spurious memories may develop. Although there no easy way determine exactly which non-word strings become...