IEEE CTCN Monthly Meeting, Wednesday, November 15, 2017: Artificial Intelligence with Search-Based Function Computation Model

#artificial #intelligence #pattern #recognition #AI #expression-based #computation #search-based #parallel
Share

The Consultants Network promotes the development of members' careers through professional and social networking, including publicizing members' skills and sharing potential professional opportunities, and supports the IEEE Central Texas Section. Website: http://ewh.ieee.org/r5/central_texas/cn/

IEEE CTCN Monthly Meeting, Wednesday, November 15, 2017: "Artificial Intelligence with Search-Based Function Computation Model"

Speaker: Chin-Liang Chang

Biography: CL Chang received his PhD in 1967 from University of California, Berkeley. He was with National Institutes of Health, IBM San Jose Research Lab, and Lockheed Software Center. He now has his own company, Nicesoft Corporation. His fields are AI, Relational Databases, and Software Engineering. He has published several books and about 50 papers. He has engaged in many discussions in the Berkeley Initiative in Soft Computing (BISC) and North America Taiwanese Engineering and Science Association (NATEA). For the more detailed information, click on the following link:

 
http://www.nicesoftsearch.com/bio.html

Abstract: We should consider search-based function computation model, not just expression-based function computation model, because if we use the search-based function computation model, then "parallel computation" can be applied. If you use the expression-based function computation model, then computations are sequential as specified in a programming language. I call the search-based function computation model "Fuzzy Similarity for Function Computation Model", which is basically a table lookup and a nearest neighbor classifier approach. For example, before we had a computer, we used "sin(x)" table. If we are given 66 degrees, we look up the table to find the result.

 
To find a "best fuzzy similar" vector, I use "parallel reduction method" by using 5000 cores in a GPU. If you have 20 million samples, divide them by 5000, each core just need to search 4000 samples. This is manageable even we need to use procedural loops.
 
In this talk, I'll first give a demo of OpenCL C++. Then, I'll describe "Fuzzy Similarity for Function Computation Model". Finally, I'll give another demo on flower image recognition.

  Date and Time

  Location

  Hosts

  Registration


  • Date: 15 Nov 2017
  • Time: 06:00 PM to 08:30 PM
  • All times are (GMT-06:00) US/Central
  • Add_To_Calendar_icon Add Event to Calendar
  • 2121 West Parmer Lane at Lamplight Village Ave.
  • Austin, Texas
  • United States 78727
  • Building: PoK-e-Jo's Smokehouse
  • Click here for Map
  • Contact Event Host

  • Bill Martino, Chairman IEEE Central Texas Consultants Network

  • Starts 27 October 2017 12:00 AM
  • Ends 15 November 2017 06:00 PM
  • All times are (GMT-06:00) US/Central
  • No Admission Charge

  Speakers

CL Chang of Nicesoft Corportation

Topic:

Artificial Intelligence with Search-Based Function Computation Model

We should consider search-based function computation model, not just expression-based function computation model, because if we use the search-based function computation model, then "parallel computation" can be applied. If you use the expression-based function computation model, then computations are sequential as specified in a programming language. I call the search-based function computation model "Fuzzy Similarity for Function Computation Model", which is basically a table lookup and a nearest neighbor classifier approach. For example, before we had a computer, we used "sin(x)" table. If we are given 66 degrees, we look up the table to find the result.

 

To find a "best fuzzy similar" vector, I use "parallel reduction method" by using 5000 cores in a GPU. If you have 20 million samples, divide them by 5000, each core just need to search 4000 samples. This is manageable even we need to use procedural loops.

 

In this talk, I'll first give a demo of OpenCL C++. Then, I'll describe "Fuzzy Similarity for Function Computation Model". Finally, I'll give another demo on flower image recognition.

Biography:

CL Chang received his PhD in 1967 from University of California, Berkeley. He was with National Institutes of Health, IBM San Jose Research Lab, and Lockheed Software Center. He now has his own company, Nicesoft Corporation. His fields are AI, Relational Databases, and Software Engineering. He has published several books and about 50 papers. He has engaged in many discussions in the Berkeley Initiative in Soft Computing (BISC) and North America Taiwanese Engineering and Science Association (NATEA). For the more detailed information, click on the following link:
 

Email:

Address:Austin, Texas, United States

CL Chang of Nicesoft Corportation

Topic:

Artificial Intelligence with Search-Based Function Computation Model

Biography:

Email:

Address:Austin, Texas, United States





Agenda

6:00 to 6:30pm -- Networking

6:30 to 8:30pm -- Business and Program


The Consultants Network meets monthly. Except when meeting jointly with other groups, the Consultants Network meets on the fourth Wednesday of each month. Meetings usually begin with informal networking from 6:00 to 6:30 p.m., followed by presentations from 6:30 to 8:30 p.m. by experts in technology, marketing, sales, advertising, financial or legal needs of small businesses and special needs of consultants.