Computer Science

The Department of Computer Science, with over 80 faculty members, has a long track record in grooming leaders for the digital economy and IT workforce. The department's internationally recognized faculty members perform research in the areas of Artificial Intelligence, Database Management, Media, Systems and Networking, Computational Biology, Programming Languages and Software Engineering. Significant results and project highlights include:

  • Development of the Effective Response Metric (ERM) model for predicting relapse of childhood lymphoblastic leukemia patients using two-time-point gene expression profiling during the first 8 days of remission-induction therapy. This novel metric is the first and (to date) only approach based on gene expression profiling that is consistently reproducible and is independently predictive of risk of relapse after adjusting for age, WBC, NCI risk, cytogenetics, and MRD.
  • Pioneering work on the Pair-End diTag (PET) idea and its associated computational analysis technologies. Many of them have been used for double-barrel shotgun sequencing for resolving ambiguous genome mapping (DNA-PET), detecting transcription factor binding sites (ChIP-PET), analysing chromatin interaction (ChIA-PET), and studying gene transcript structure (RNA-PET), etc.
  • Tracer: a verification tool based on symbolic execution with "Lazy Annotations". The general technique is to formulate program analysis as a combinatorial optimization problem, and uses dynamic programming in symbolic search. TRACER system has been used for many applications, such as timing analyses, slicing, testing, and concurrency optimization. This general approach is a foundation that holds big promise for solving the difficult problem of providing a practical and uniform basis for analyzing, verifying, and testing computer programs both statically and dynamically.
  • SINGA: A General Distributed Deep Learning Platform SINGA is an Apache Incubator Project. It is a general distributed deep learning platform for training big deep learning models over large datasets. The key goals of SINGA is usability, scalability and extensibility. SINGA offers a simple and intuitive programming model, making it accessible to even non-experts. SINGA provides a general architecture to exploit the scalability of different training frameworks. Synchronous training frameworks improve the efficiency of one training iteration, and asynchronous training frameworks improve the convergence rate. Given a fixed budget (e.g., cluster size), users can run a hybrid framework that maximizes the scalability by trading off between efficiency and convergence rate. SINGA is extensible and able to support a wide range of applications requiring different deep learning models.
  • RetinaCloud: Platform for the detection, screening and management of major eye diseases, and study of systemic diseases. It allows for the documentation of retinal structural alterations, and monitor these changes over time. The platform currently support two flagship systems: SIVA and SELENA. SIVA extracts retinal blood vessel structure and derives a spectrum of quantitative measurements including caliber, tortuosity, bifurcation and fractals. It is able to attain the desired clinical sensitivity for large-scale population studies to be conducted, and greatly reduces the manpower requirements for analyzing retinal images. SELENA focuses on the detection of lesions, particular the lesions of diabetic retinopathy, in retinal images. This system optimizes the trade-off between high sensitivity and specificity to automatically screen patients for diabetic retinopathy at the polyclinics and other primary care providers, leading to a considerable reduction in workload in grading by family physicians, increased efficiency and potentially, substantial savings to the Singapore health system and patients.
  • 1
  • 2
  • 3
  • 4
  • 5