- Data Quality and Management
- Network Security and Intrusion Detection
- Neural Networks and Applications
- Anomaly Detection Techniques and Applications
- Ferroelectric and Negative Capacitance Devices
- Privacy-Preserving Technologies in Data
- Cognitive Science and Mapping
- Semantic Web and Ontologies
- Neural dynamics and brain function
- Functional Brain Connectivity Studies
- Advanced Database Systems and Queries
- Web Data Mining and Analysis
- Imbalanced Data Classification Techniques
- Financial Distress and Bankruptcy Prediction
- Caching and Content Delivery
- Data Management and Algorithms
- Topic Modeling
- Action Observation and Synchronization
- Advanced Memory and Neural Computing
- Evolutionary Algorithms and Applications
- Language, Metaphor, and Cognition
- Data Stream Mining Techniques
- Spam and Phishing Detection
- Computability, Logic, AI Algorithms
- Neurobiology of Language and Bilingualism
La Trobe University
2005-2014
Johns Hopkins University
1993
Jackendoff (2002) posed four challenges that linguistic combinatoriality and rules of language present to theories brain function. The essence these problems is the question how neurally instantiate rapid construction transformation compositional structures are typically taken be domain symbolic processing. He contended typical connectionist approaches fail meet dialogue between theory cognitive neuroscience will relatively unproductive until importance widely recognised answered by some...
Identity crime is well known, prevalent, and costly; credit application fraud a specific case of identity crime. The existing nondata mining detection system business rules scorecards, known matching have limitations. To address these limitations combat in real time, this paper proposes new multilayered complemented with two additional layers: communal (CD) spike (SD). CD finds social relationships to reduce the suspicion score, tamper resistant synthetic relationships. It whitelist-oriented...
Entity resolution, also known as data matching or record linkage, is the task of identifying and records from several databases that refer to same entities. Traditionally, entity resolution has been applied in batch-mode on static databases. However, many organisations are increasingly faced with challenge having large containing entities need be matched real-time a stream query entities, such best retrieved. Example applications include online law enforcement national security databases,...
A Bloom filter is a simple data structure supporting membership queries on set. The standard does not support the delete operation, therefore, many applications use counting to enable deletion. This paper proposes generalization of approach, called "autoscaling filters", which allows adjustment its capacity with probabilistic bounds false positives and true positives. In essence, autoscaling binarized an adjustable binarization threshold. We present mathematical analysis performance as well...
Forty pneumonia and 40 fracture cases were matched by patient age gender with normal cases. Pediatric constituted 25% of this sample. All 160 examinations laser-digitized at a spot size 210 microns, contrast resolution 12 bits, spatial 2.35 line pairs per millimeter. The digitized images transmitted 5 miles over dedicated telephone T1 rate from Francis Scott Key Hospital to Johns Hopkins (both in Baltimore, Md). At Hospital, eight radiologists interpreted the original analog radiograph...
Analogy-making is a key function of human cognition. Therefore, the development computational models analogy that automatically learn from examples can lead to significant advances in cognitive systems. Analogies require complex, relational representations learned structures, which challenging for both symbolic and neurally inspired models. Vector architectures (VSAs) are class connectionist representation manipulation compositional be used model analogy. We study novel VSA network...
Real-time Entity Resolution (ER) is the process of matching query records in subsecond time with a database that represent same real-world entity. Indexing techniques are generally used to efficiently extract set candidate from similar record, and be compared record more detail. The sorted neighborhood indexing method, which sorts compares within sliding window, has been successfully for ER large static databases. However, because it based on arrays designed batch resolves all rather than...
This paper proposes a simple encoding scheme for words using principles of Vector Symbolic Architectures. The proposed allows finding valid word in the dictionary given permuted (represented approach) only single operation– calculation Hamming distance to distributed representations dictionary. can be used as an additional processing mechanism models embedding, which also form vectors represent meanings words, order match distorted text
Introduction. Neural network models of autoassociative, distributed memory allow storage and retrieval many items (vectors) where the number stored can exceed vector dimension (the neurons in network). This opens possibility a sublinear time search (in items) for approximate nearest neighbors among vectors high dimension. The purpose this paper is to review that be naturally implemented by neural networks (mainly with local learning rules iterative dynamics based on information locally...
The authors, on the basis of brief arguments, have dismissed tensor networks as a viable response to Jackendoff's challenges. However, there are reasons believe that connectionist approaches descended from actually very well suited answering I rebut their arguments for dismissing and briefly compare approaches.
Textual databases are ubiquitous in many application domains. Examples of textual data range from names and addresses customers to social media posts bibliographic records. With online services, individuals increasingly required enter their personal details for example when purchasing products or registering government while network e-commerce sites allow users post short comments. Many leave open the possibility people unintended malicious abnormal values, such as with errors, bogus profane...
This paper is on adaptive real-time searching of credit application data streams for identity crime with many search parameters. Specifically, we concentrated handling our domain-specific adversarial activity problem the Communal Analysis Suspicion Scoring (CASS) algorithm. CASS's main novel theoretical contribution in formulation State-of- Alert (SoA) which sets condition reduced, same, or heightened watchfulness; and Parameter-of-Change (PoC) improves detection ability pre-defined...