LESSA: a Library of Building Blocks for Efficient, Scalable and Service-oriented Algorithms

1.Machine Learning Foundation Techniques and Applications

We are dedicated to developing foundation techinques in machine learning, including but not limited to foundation models (e.g. large large models and large graph models), gradient desecent optimizers, distrbuted learning systems (e.g. deep learning and graph learning), deep learning compilers and automatic machine learning (AutoML), and further develop practical applications to solve real-world problems.

Example work:

2. Compact Computing for Big Data

As data volume increases exponentially and data compression is popularly used in data centers, I would expect that directly operating on compressed data would become commonplace in the future. However, parallel processing of compressed data is a challenging proposition for both shared-memory and distributed-memory systems and the challenges could come from constrained random accesses to uncompressed content, independent decompression of data blocks, balanced distribution of data, memory and computation, adaption of existing algorithms and applications to meet the requirements of compressed data processing and so on. Based on these concerns, I am conceiving of a new concept of computing and name it as Compact Computing tentatively. In general, Compact Computing targets robust, flexible and reproducible parallel processing of big data and consists of three core components, in principle: (1) tightly-coupled architectures, (2) compressive and elastic data representation, and (3) efficient, scalable and service-oriented algorithms and applications. In this context, component 1 can comprise conventional CPUs, a diversity of accelerators (e.g. FPGAs, GPUs, MIC processors and etc.) and fast interconnect communication facilities; component 2 concentrates on data structures and formats that enable efficient on-the-fly streaming compression and decompression; and component 3 targets the development of algorithms and applications that enable robust and efficient processing of big data streams in parallel. By centering around data, Compact Computing enables full-stack computation by tightly coupling algorithms with systems.

Example work:

3. Large-Scale Biological Sequence Analysis System

As part of my LESSA library, this project has been the core of my research on parallel and distributed algorithm design for bioinformatics, by employing a variety of tightly-coupled and loosely-coupled computing architectures, including heterogeneous computers with accelerators (e.g. Intel SSE, Intel AVX, Intel Xeon Phis, NVIDIA GPUs and AMD GPUs), cluster computing and cloud computing. My final objective is to establish an analysis system for large-scale biological sequences, in order to solve some critical and bottleneck problems in bioinformatics and computational biology, such as genome sequencing based on high-throughput sequencing technologies, meta-genomics, motif discovery, sequence alignment, and phylogenetic inference. Fig. 1 illustrates the diagram of an imaginaory biological data analysis system.

Fig. 1 System diagram for large-scale biological sequence analysis

Example work:

(The full lists of my software and publications are available here)