Lung-Pin Chen

Software Development Framework for Job-Level Algorithms

Recently, Wu et al. introduced a general distributed computing approach, named Job-Level (JL) Computing. In JL computing, a search tree is maintained by a client process, while search tree nodes are evaluated, expanded, or generated by leveraging game-playing programs. These node operations are encapsulated as coarse-grain jobs, each requiring tens of seconds or more of running the programs. This article presents an abstraction of the JL computing approach and develops a general JL search framework so that common modules may be reused for various JL applications, making JL development easier. We describe in detail the implementation of the JL Proof-number Search (JL-PNS) and JL Upper Confidence Bound Search (JL-UCT) as case studies in the application of the JL search framework. In our case studies, only hundreds of lines of code are required for new JL applications, while the code for the JL framework consists of more than ten thousand lines of code. It demonstrates that this framework can be used to greatly reduce new JL application development and software maintenance efforts.

Efficient online algorithm for identifying useless states in distributed systems

In a distributed system, detecting whether a given logical predicate is true on the global states is fundamental for testing and debugging the program. Detecting predicates by examining all global states is intractable due to the combinatorial nature of the problem. This work designs an efficient online algorithm that identifies the consistent and useless states each time a new state is reported. This paper formulates the optimality of detecting algorithms in terms of pseudo states, which are employed to represent unknown states to the monitor process. Based on this technique, memory space of the debugger can be minimized by removing the useless states without affecting the debugging results. While minimizing memory space, the proposed algorithm requires only O(p2M) time in total, where p is the number of processes, and M is the number of reported states.

Software Framework for Generic Game Development in CGDG

In this paper, we propose a software framework for easily facilitating computer game developments. Using the software framework, the developers can easily implement game record editors for computer game AI analysis. Additionally, using the framework, the developers can easily design job-level (JL) search algorithm to dispatch jobs to workers of a desktop grid, named CGDG. Based on the framework, we successfully developed several record editors for many games, such as Connect6, Go, Chinese Chess, Mahjong, and also developed some JL search algorithms, such as JL proof number search (JL-PNS), JL Monte-Carlo tree search (JL-MCTS), JL alpha-beta search (JL-ABS). Most importantly, once a JL algorithm such as JL-PNS is implemented, it can be applied to all games. This demonstrates the excellence of this framework.

A Loosely Coupled Interactive Web Data Extraction System

As the rapid growing of Internet, the Web data extraction (DE) system has become a convenient tool for application programs to collect useful data. A DE system takes a wrapper as its input, which is a script describes how to navigate Web pages and extract the data. The integration of application program, wrappers, and DE system is a nontrivial task due to the complete coordination and interaction among application programs and DE systems. Furthermore, the asynchronous update technologies used in many Web pages, such as AJAX, make the integration more complex. This paper proposes a loosely coupled interactive DE system based on Browser-Oriented Data Extraction (BODE) systems and the Web Service Resource Framework (WSRF), a web service specification which standardizes the accessing and manipulation of states for web services. The loosely coupled interactive DE system provides users with the states of wrappers through the web service states, called WS-Resources. The interactive DE system can also accept parameters from the application programs during extraction through the WSRFs notification design pattern. By providing the above interactive capabilities, the application programs and wrappers of the DE systems can be easily shared and controlled.

Enhancing parallel game-tree searches by using idle resources of a high performance render farm

Due to the complex scheduling criteria, resource fragmentation often occurs in modern high performance computing systems. This paper addresses both of the policy and technical issues of using unstable resources formed by fragmentation. We propose that the parallel Monte-Carlo tree search (MCTS) can be implemented as lightweight single-core tasks to adapt to the unstable fractured resources. We develop a broker to collect idle resources in a high performance render farm and make use of them to execute MCTS tasks. The broker tends to assign promising tree nodes to reliable and responsive processors to confirm best-first search. Our work demonstrates a successful integration in which the MCTS computation gains significant resources without interfering with the render farm tasks.

Efficient Deployment of Content Applications in a Cloud Environment

For many Internet applications, the contents are built based on multiple data sources that change continually, and yet the presentation formats have to be tailored according to various user profiles. This paper investigates the dynamic content applications that are modeled as workflows of XML objects. Each content object can be constructed from its predecessors by executing the transformation scripts such as XSLT or XQuery scripts. This paper develops an algorithm for deploying content objects to a cloud platform subjected to minimize the CPU time as well as the network bandwidth used by servers. By transforming to the well-studied graph cut problem, the proposed algorithm solves the problem efficiently in polynomial time.