Saleh Elmohamed

NPAC Grad Student Agreement Form || Saleh Elmohamed

Date:

October, 1996.

Mailing Address:

On campus: 111 College Place, Rm 3-217, NPAC CST Bld. Syracuse, NY 13244-4100
Off campus: P.O. Box 747, Univ. Station, Syracuse, NY 13210-0747

Email address:

saleh@npac.syr.edu

SU Academic Department affiliation:

Affiliated with School of Computer and Information Science .

Degrees earned:

B. Science in Math-Computer Science (Alexandria, Egypt and Michigan)
Postgrad work in Applied Mathematics (Oregon State Univ.)
Sci. Master in Computer and Information Sci. (Syracuse Univ.)

Current Academic Status at SU:

PhD student

Date Course Work Started:

1993 (after the master)

Anticipated Comprehensive Examination Date:

Anticipated Masters graduation Date:

1993-94

Anticipated PhD graduation Date:

No later than May 1997.

Total Number of Tuition Hours to Complete Current Degree:

45 (out of this 20 hours or so for the thesis and the rest are course work and independent studies).

Number of Tuition Hours needed for Fall:

6 credit hours.

Number of Tuition Hours needed for Spring:

Number of Tuition Hours needed for Summer:

Name of NPAC Project leader:

Paul Coddington

Full Name of Academic Advisor:

Geoffrey C. Fox

Full Name of Thesis/Dissertation Director:

Geoffrey C. Fox

Full Names of Thesis/Dissertation Committee Members:

Geoffrey C. Fox
Paul Coddington
?
?
?

List all Completed SU Courses and Grade Earned:

CIS-626 (theoretical found. of computer science) [A]
ECE-690 (indpt. study; distributed systems) [A]
CIS-600 (topic: expert database systems) [A]
CIS-600 (topic: frontiers in logic programming) [A]
CIS-500 (artificial intelligence) [A]
CIS-690 (topic: data parallel programming on the CM) [A-]
CIS-623 (fund. of programming) [A]
CIS-690 (topic: neural networks) [A]
CIS-660 logic programming and optimization [A]
CIS-672 mathematical logic [A-]
ECE-791 special probability prin. of expert systems [B+]
CIS-662 logic programming II [B]
CIS-700 natural language processing [B]
CPS-616 web technology and Java programming
CPS-713 case studies in computational science (optimization) [A-]
CPS-615 computational science [A]
NEU-900 issues in Neuro-science [A-]
MTH-487 (OSU course) numerical methods for scientific applications [A]
MTH-505 (OSU course) adv. numerical methods for scientific app. [A]
CS-516 (OSU course) adv. topics in op. systems [B]
CS-562 (OSU course) adv. software engineering methods [A]
CS-551 (OSU course) algorithm theory and application [A]
CS-553 (OSU course) algorithm theory and computational complexity [A]
CS-564 (OSU course) database systems [A-]
CS-569 (OSU course) software systems [A]

List all SU Courses you EXPECT to Take (for fall 1996):

CIS-600 Topics in AI (Modeling) CIS-690 Indep. Study (HPSE optimization - schedeling, etc.)

List your PhD Reading List:

(NA), but will involve lots of material on heuristics for solving combinatorial optimization problems.

Dissertation Topic:

Combinatorial Optimization Methods and Heuristics for Tackling non-Euclidean and Constraint Satisfaction type Problems (e.g. scheduling and timetabling)

Career Goals:

A position in academia or a research lab/Organization will be most desirable. Since I received most of my education here in the US, it is only fitting to have the preference to serve and work in the USA.

Areas of your specialization:

Areas of Knowledge and Interest
- Scheduling, Timetabling, and Graph Coloring
- Heuristics and Algorithms for Optimization
- Geometric Modeling, System Simulation and Modeling, and Object-Oriented Software Analysis
- Scientific Computation, Modeling, and Simulation on both SIMD and MIMD architectures
- Associative Memory Models, Visual Learning Tools, and Data Analysis
- Neural Network Models (forward and recurrent): Simulation and Theoretical Analysis
- Software Engineering (planning, designing, testing, and analysis)
Software/Hardware Systems and Languages
- Programming Skills: proficient in Fortran, C, Pascal, Common Lisp, Prolog, C++, Snobol, and Cobol
- OS Experience: Unix OS, VMS, CMS, MS DOS, and SunOS
- Hardware Systems: Have experience in the following systems: SunSparcs, Vax 780s and 750s, Tektronix stations, DecStations, HP stations, Intel iPSC/860, BBN Butterfly, IBM 3090 and PCs, Apple Macs, CM-2 and CM-5, and IBM SP-2
- Some of the medium to large scale projects I have done
  - Scheduling and timetabling using quite large real world data. All simulations were written in C with various shells. Computing platforms used are SUNs and the IBM SP2 and RS-6000s.
    - S. Elmohamed, G. Fox, and P. Coddington simulated annealing and non-Euclidean combinatorial optimization problems: the case of academic course scheduling, theory and implementation , to be published, 1996. (also as SCCS report)
    - S. Elmohamed, G. Fox, P. Coddington mean field annealing and non-Euclidean combinatorial optimization problems: the case of academic course scheduling, theory and implementation , to be published, 1996. (also as SCCS report)
  - Neural Networks (recurrent and feedforward) simulation on SUNs (coded in C). The same type of simulations were carried out on the CM-2 and later on the CM-5. This particular work involves testing and simulation of various neural net topologies on the Connection Machine. The variation in the topology is in (1) the number of nodes and layers (2) the type of neuron function (e.g. sigmoid, quadratic, linear, piecewise linear, rational, Gaussian, etc.) (3) the I/O mapping, and (4) the net's connectivity. For more details, see the following reports:
    - S. Elmohamed, classification and classifiers: neural vs. conventional methods , unpublished report, 1993.
    - S. Elmohamed, Some issues of reducibility and equivalence in feedforward neural networks , SCCS report, Jan 1996.
    - S. Elmohamed, An evaluative study of an alternative to back-propagation training algorithm for neural networks , SCCS report?, Jan 1996.
  - Temporal knowledge-based system was developed and simulated on SunSparcs (coded in Prolog and C).
  - While I was with OSU, a set of communication protocols were designed and implemented on HP-9836 workstations and some old VAX-750. The purpose of such protocols is to establish line and forms of communication between the 2 systems allowing the users at both ends to either broadcast or use direct addressing to send/receive messages and software. Those communication tools were implemented in C. In addition, the second phase of the above communication project consists of porting prototypes of Modula-2 and Edison systems between the aforementioned systems. This was a joint work with Prof. T. Minura , CS dept, Oregon State Univ.
    - Elmohamed and Minura design and implementation of communication protocols between HP-9836 and Vax-750s , Applied Math and CS report CS/MAT-AM-84/22.
    - Elmohamed and Minura software porting between Vax and HP-9836 workstations , Applied Math and CS report CS/MAT-AM-84/27.
  - Modeling and simulation of physical systems in engineering and biology. This work was part of my research in applied math supervised by Profs. Ronald Guenther and Joel Davis (at Oregon State). This work involved extensive computer modeling and simulation of physical and engineering science problems. The simulators utilized a host of numerical techniques and algorithms with the end goals of optimization, time/space efficiency and accuracy. The set of numerical methods includes: finite differences, interpolation, integration, linear systems, polynomials, differential equations, and Fourier and Laplace transforms and functions. The systems we modeled and simulated were (1) From chemical engineering: a network of chemical processes, (2) From ecology: a small ecosystem of pollution and its effects on human population growth, and vice verse. These simulations were written in C and Fortran.
Project assignments:
- Putting together an HPSE "roadmap" to the numerous tools and resources on the net that utilizes neural networks, genetic algorithms, simulated annealing, and mean field annealing for tackling useful optimization problems (such as graph matching/partition, scheduling, timetabling, etc.). When completed, this roadmap is supposed to be part of the existing resources on optimization that reside on the local web.
- Also as part of my fall 96 support is to assist Geoffrey with CPS-615 by grading, setting up homework assignments and assisting students in using the virtual programming lab (VPL).
On what NPAC Main project(s) are you currently working:
I am working on the following:
- Worked on scheduling and timetabling using data from the scheduling office of Syracuse University. The methods used are:
  - Expert systems (rule-based).
  - Simulated annealing, using an expert system as a preprocessor.
  - Mean field annealing (a recurrent-net or a Hopfield-type net with mean field theory equations to control its dynamics).
  - Mean field annealing, also using an expert system as a preprocessor.
  The results obtained (in particular those of simulated annealing with the expert system acting as a preprocessor) are quite encouraging and far better than those obtained using the current methods used by Syracuse University. We will report on this shortly.
- Since the beginning of summer'95, Paul and I have been working on a different approach to the problem. In this approach we map the timetabling onto graph coloring then tackle the latter one to obtain a solution for the former problem. This is quite intuitive, and from a theoretical point of view is always the recommended approach when tackling this kind of non-Euclidean combinatorial optimization problems. So far, the results obtained are much better than those of the expert system approach. We (Hon, Paul, and I) will continue working on timetabling using the graph coloring approach taking into account most (if not all) of the soft and medium constraints as well as the hard ones.
Name of associated NPAC project leader(s):
- Paul Coddington
On what NPAC Service activities are you currently working:
- Will continue on the same line of work done over the summer but at a different level and that is to deal with most of (if not all of) the constraints involved and mapped them onto graph coloring.
- Substitute graph coloring and its results for those obtained from the expert system then use those results as an input to simulated annealing in the goal of obtaining a much better results. One other option is not to try to deal with all of the constraints at the preprocessor level but leave some to be handled during the annealing process.
- Compare and analyze results before and after the substitution,
- Incorporate this approach with others and make available on the web as part of the aforementioned "roadmap" .
- Wrap up the neural nets and simulated annealing papers and submit them for a review.
- Grading and assisting students with their work for CPS-615.
Name of associated NPAC project leaders:
- Paul Coddington
List other NPAC activities you have worked on in the past which are not covered above:
All covered above.
Name of associated NPAC project leaders:
Paul Coddington
Type of research you expect to be working on for the next semester:
Also will continue to work on the following subtasks:
- Further research for new and more novel heuristics to deal with academic scheduling and timetabling in general. Testing these methods on large data sets, in the presence of many non-trivial hard and soft constraints, is necessary to determine the viability and efficiency of such heuristics.
- Continue to use simulated annealing with adaptive and smart cooling schedules to deal with the abovementioned problem.
- Carry on with graph coloring (GC) line of investigation to determine whether a much larger set of constraints, than the one we have tried, is mappable to GC. If such a set is mappable then work out the implementation and compare results.
- In regard to graph coloring, dealing with a full set of timetabling constraints will definitely lead to denser graphs, consequently we will need to use a different and a more suitable heuristics for such graphs than the one used (the saturation heuristics) for sparse graphs.
- Carry on with the web work incorporating any other novel approaches to the current information.
  Any expected absences during the academic year:
  NA
  Approvals:
  
  NPAC Project Leader:
  Paul Coddington
  Thesis Dissertation Director:
  Geoffrey C. Fox
  Academic Advisor:
  Geoffrey C. Fox
  NPAC Director:
  Geoffrey C. Fox