Vittorio Castelli's Research Interests

Personal Wizards Capturing procedural knowledge from demonstrations. Constructing executable models of procedures from multiple demonstrations. Automatically generating documentation by observation. Incremental, real-time, learning of procedure models Information Theory Properties of dictionary-based lossless compression schemes, such as LZ Achievable regions for the Multiple Description Problem Autonomic Computing Software Rejuvenation: prediction of medium-term resource exhaustion (developed algorithms used in IBM Director)  Capacity Management: prediction of long-term resource utilization in computer systems (developed algorithms  used in IBM Director). MXT: Memory Xpansion Technology Memory Hierarchy Organization of computer systems that keep the content of main memory in compressed form. End-to-End Compression: architecture of computere systems where data and programs are stored, transmitted, and transferred between different architectural components in compressed form. Dictionary-Based Compression: properties of dictionary-based compression algorithms, including parallel LZ. Digital Libraries and Content-Based Retrieval Content-based retrieval from databased of scientific data: image processing on compressed data;  statistical classification of remotely-sensed images; definition and execution of complex, multimodal queries on large image databases. Application of remotely-sensed data to public health.  Multidimensional indexing structures Algorithms for approximate similarity queries in high-dimensional spaces. Indexing structures for similarity queries using flexible scoring functions. Relevance feedback for similarity retrieval.

Last Modified on 04/29/2004