On%20Some%20Characteristics%20of%20the%20New%20Generation%20Computers - PowerPoint PPT Presentation

About This Presentation
Title:

On%20Some%20Characteristics%20of%20the%20New%20Generation%20Computers

Description:

(sysoev_s_at_mail.ru) The Computer Generations. 1. 4. 2. 3. 5. Qualitative Generations Leaps ... For each considered particular task the system has the particular ... – PowerPoint PPT presentation

Number of Views:52
Avg rating:3.0/5.0
Slides: 17
Provided by: ser8181
Category:

less

Transcript and Presenter's Notes

Title: On%20Some%20Characteristics%20of%20the%20New%20Generation%20Computers


1
On Some Characteristics of the New Generation
Computers
  • Oleg N. Granichin (oleg_granitchin_at_mail.ru)
  • Sergey S. Sysoev
  • (sysoev_s_at_mail.ru)

2
The Computer Generations
1
2
3
4
?
5
3
Qualitative Generations Leaps
Task oriented machines.
Operating Systems, multitasking, Universal
Computer, HLL
4
What Is Next?
The Intelligent System
5
One Approach of Creating the Intelligent System
The Internal Part
The External Part
6
Intelligent Choice

Task 1
Task 2
Task N

Device N
Device 1
Device 2
7
The Internal Part Conditions
  • For each considered particular task the system
    has the particular device which is able to
    optimize the solving of this task by an
    appropriate choosing the system parameter from
    some final set.
  • Information from external world must be delivered
    to all such devices simultaneously.

8
The Internal Part Implementation Example
9
Informational Resonance
10
Possible Resonance Situations
  • One and only one device has resonance
  • Several devices have resonance
  • None of the devices has resonance (the system
    parameters have to be adjusted)

11
The System Parameters Set
12
Optimization Task
The function F has its minimum (or maximum) when
the system has resonance, x the system
parameters, w the stochastic parameter,
obtained from the real world.
13
The Kiefer-Wolfowitz Procedure
14
The SPSA Algorithm
15
The SPSA Grounds
16
Advantages of the SPSA Algorithm
  • Only one measurement of loss function per
    iteration needed
  • Convergence with probability 1 with all kinds of
    disturbance
  • The best convergence speed characteristics
  • Allows tracking
Write a Comment
User Comments (0)
About PowerShow.com