"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
The Cognitive Net is Coming
1. The Cognitive Net is
Coming
Liotta,A
IEEE Spectrum, Vol. 50, No. 8, p.26-.
ISSN 0018-9235
August 2013
2. Introduction
Major Internet service providers around the world are
now reporting global latencies greater than 120
milliseconds
How slowly traffic would move if console gamers and
cable television watchers, who now consume hundreds
of exabytes of data off-line, suddenly migrated to cloud–
based services.
The problem is not simply one of volume
Network operators will always be able to add capacity by transmitting
data more efficiently and by rolling out more cables and cellular base
station.
The real troubles lies with the technology at the heart of the internet is its
routing architecture
4. The Path to Intelligent Routing
The future Internet will need smarter routing algorithms
to handle diverse data flows and prevent failures
Although there are no tried and true solutions yet, early
designs might follow an architecture like this one
Routing Device
Can be any network node, such as a phone, a television, a
car or an environment sensor
The Routing and Forwarding Engines
Determines the best pathways to get data packets to their
destinations and queue them for transmission
5. The Path to Intelligent Routing (Cont’d)
The autonomic controller
Directs the routing and forwarding engines by following the MAPE
loop: It monitors internal sensor data and signals from other nodes,
analyzes that information, plans a response, and executes it.
The cognitive engine
Helps the router adapt to unforeseen changes by following the
OOPDAL loop: It observes the environment, orients the systems by
prioritizing tasks, plans options, decides on a plan, acts on it and
learns from its action.
7. Monitor-Analyze-Plan-Execute
(MAPE) loop
One idea, proposed by IBM
Algorithm of this architecture must perform four main
task:
They monitor a router’s environment
battery level,
memory capacity,
the type of traffic it’s seeing,
the number of nodes it’s connected to, and
the bandwidth of those connection
The knowledge algorithms analyze all that data
Use statistical techniques to determine whether the inputs are typical and if
they aren’t, whether the router can handle them
Example: if the router that typically receives low-quality video streams
suddenly receives a high-quality one, the algorithms calculate whether the
router can process the stream before the video packers fill its buffer
8. Monitor-Analyze-Plan-Execute
(MAPE) loop (Cont’d)
They plan a response to any potential problem
Such as an incoming video stream that’s too large
Example :They may figure the best plan is to ask the
video server to lower the stream’s bit rate. Or they may
find its better to break up the stream and work with other
nodes to spread the data over many different pathways.
They execute the plan
The execution commands may modify the routing tables,
tweak the queuing methods, reduce transmission power,
or select a different transmission channel, among many
possible actions
9. The cognitive engine
This architecture used cognitive algorithms
Unlike autonomic system, which rely on
predetermined rules, cognitive algorithm make
decisions based on experience
Example :When you reach for a ball flying toward you,
for example, you decide where to position your hand
by recalling previous successes. If you catch the ball,
the experience reinforces your reasoning. If you drop
the ball, you will receive your strategy
10. The cognitive engine (Cont’d)
Cognition algorithms orient the system by evaluating and
prioritizing the gathered information
For low-priority actions, the algorithms consider alternative
plans. Then they decide on a plan and act on it, either by
triggering new internal behavior or by signaling nearby nodes
When more-urgent action is needed, the algorithms can
bypass one or both of the planning and decision-making steps
Finally, by observing the results of these actions, they would
generate prediction models that would continually modify the
knowledge algorithm, thereby improving the router’s ability to
manage diverse data flows
Editor's Notes
Title ideas:
(was) Cognitive and Information Science
Others:
The Cognitive Perspective in Information Science Research
Cognitive Science and the Cognitive Perspective in IS: Background and Applications
??...
Not discrete areas that can be studied without touching on the other areas.
In LIS, ideas from each of these areas apply in overlapping and interconnecting ways. For example, in the process of building intelligent agents.
Breakdown of general topic areas from:
Eysenck, M.W. ed. (1990). The Blackwell Dictionary of Cognitive Psychology. Cambridge, Massachusetts: Basil Blackwell Ltd.
Definition:
Perception is the complex sequence of processes by which we take the information received from our senses and then organize and interpret it, which in turn allows us to see and hear the world around us as meaningful, recognizable objects and events with clear locations in space and time. (encyl. Of Cog Sci)
Perception:
Is limited
Is selective
Refers to the distal stimulus, not the proximal stimulus (the physical object or event, not the light/sound/whatever patterns that arrive at sensory receptors)
Requires time
Not entirely veridical (accurate, faithful, or with high fidelity)
Requires memory
Requires internal representations
Is influenced by context
Definition:
The mental manipulation of symbolic representations of objects and of relations among objects
Development of taxonomy of thinking skills:
Verbal reasoning
Argument analysis
Thinking as hypothesis testing
Likelihood and uncertainty
Decision-making and problem-solving
Reasoning
-Thought as formal logic. Inferring AI, etc.
Stages of problem solving
Mental models
Stone, D J The influence of mental models and goals on search patterns during Web interaction Journal of the American Society for Information Science and Technology; 53 (13) 2002, p.1152-1169