Adel Ardalan

In-RDBMS hardware acceleration of advanced analytics

The data revolution is fueled by advances in several areas, including databases, high-performance computer architecture, and machine learning. Although timely, there is a void of solutions that brings these disjoint directions together. This paper sets out to be the initial step towards such a union. The aim is to devise a solution for the in-Database Acceleration of Advanced Analytics (DAnA). DAnA empowers database users to leap beyond traditional data summarization techniques and seamlessly utilize hardware-accelerated machine learning. Deploying specialized hardware, such as FPGAs, for in-database analytics currently requires hand-designing the hardware and manually routing the data. Instead, DAnA automatically maps a high-level specification of in-database analytics queries to the FPGA accelerator. The accelerator implementation is generated from a User Defined Function (UDF), expressed as part of a SQL query in a Python-embedded Domain Specific Language (DSL). To realize efficient in-database integration, DAnA accelerators contain a novel hardware structure, Striders, that directly interface with the buffer pool of the database. DAnA obtains the schema and page layout information from the database catalog to configure the Striders. In turn, Striders extract, cleanse, and process the training data tuples, which are consumed by a multi-threaded FPGA engine that executes the analytics algorithm. We integrated DAnA with PostgreSQL to generate hardware accelerators for a range of real-world and synthetic datasets running diverse ML algorithms. Results show that DAnA-enhanced PostgreSQL provides, on average, 11.3x end-to-end speedup than MADLib and 5.4x faster than multi-threaded MADLib running on Greenplum. DAnA provides these benefits while hiding the complexity of hardware design from data scientists and allowing them to express the algorithm in 30-60 lines of Python.

Mathematical Analysis and Computational Integration of Massive Heterogeneous Data from the Human Retina

Modern epidemiology integrates knowledge from heterogeneous collections of data consisting of numerical, descriptive and imaging. Large-scale epidemiological studies use sophisticated statistical analysis, mathematical models using differential equations and versatile analytic tools that handle numerical data. In contrast, knowledge extraction from images and descriptive information in the form of text and diagrams remain a challenge for most fields, in particular, for diseases of the eye. In this article we provide a roadmap towards extraction of knowledge from text and images with focus on forthcoming applications to epidemiological investigation of retinal diseases, especially from existing massive heterogeneous collections of data distributed around the globe.

Design and Applications of Intelligent Systems in Identifying Future Occurrence of Tuberculosis Infection in Population at Risk

Tuberculosis is a treatable but severe disease caused by Mycobacterium tuberculosis (Mtb). Recent statistics by international health organizations estimate the Mtb exposure to have reached over two billion individuals. Delay in disease diagnosis could be fatal, especially to the population at risk, such as individuals with compromised immune systems. Intelligent decision systems (IDS) provide a promising tool to expedite discovery of biomarkers, and to boost their impact on earlier prediction of the likelihood of the disease onset. A novel IDS (iTB) is designed that integrates results from molecular medicine and systems biology of Mtb infection to estimate model parameters for prediction of the dynamics of the gene networks in Mtb-infected laboratory animals. The mouse model identifies a number of genes whose expressions could be significantly altered during the TB activation. Among them, a much smaller number of the most informative genes for prediction of the onset of TB are selected using a modified version of Empirical Risk Minimization as in Vapnik’s statistical learning theory. A hybrid intelligent system is designed to take as input the mRNA abundance at a near genome-size from the individual-to-be-tested, measured 3-4 times. The algorithms determine if that individual is at risk of the onset of the disease based on our current analysis of mRNA data, and to predict the values of the biomarkers for a future period (of up to 60 days for mice; this may differ for humans). An early warning sign allows conducting gene expression analysis during the activation which aims to find key genes that are expressed. With rapid advances in low-cost genome-based diagnosis, this IDS architecture provides a promising platform to advance Personalized Health Care based on sequencing the genome and microarray analysis of samples obtained from individuals at risk. The novelty of the design of iTB lies in the integration of the IDS design principles and the solution of the biological problems hand-in-hand, so as to provide an AI framework for biologically better-targeted personalized prevention/treatment for the high-risk groups. The iTB design applies in more generality, and provides the potential for extension of our AI-approach to personalized-medicine to prevent other public health pandemics.