skaisler’s Site

If you decide you want to develop an ML system, you need to decide on how to implement the component's of Scott's model.

Representations for data and knowledge come in many forms. We have talked about a few of those over the past two weeks. The representation should be a natural fit for the domain data and the domain knowledge. As we noted in lecture 4, there are both quantitative and qualitative approaches, e.g., the numbers vs. symbols dichotomy, and the possibility of a hybrid approach.

Some representations include: decision trees, version spaces (Mitchell), kNN classification, rules of various types, first order logic statements, graphs/networks, blackboard systems, tuples (such as RDF and XML), etc.

For the memory, you need to consider both a working or transient memory as well as a persistent memory. Working memory often holds temporary or intermediate results, while persistent memory holds accumulated knowledge including the solutions to specified problems.

G is the event generator that creates the inputs to the learning procedure. G may be internal or external to the MLS. in school, teachers represent external generators to the students - at least through high school and, perhaps, parts of college. For a sensor system, say a radar, it could be the periodic pulse-and-return event of the radar system. The return is the event that the MLS would see as an input and process. An internal example is a kmeans clusterer or a neural network. Both receive an initial external event, but subsequent events are the outcomes of applying the event generation procedure(s) internally to the system For kmeans, it is the succeeding recomputation of the clusters, while for NN, it is the results of one pass through the network.

The learning procedure P is the mechanism by which you system "learns" (if it does - see the initial sentence in the previous blog. P, which may consist of multiple functional mechanisms, constructs or modifies the representations in memory (either one or both). In effect, P encompasses a set of transformations. For example, one iteration of a neural network where the representations is the values and weights at each layer of that iteration.

The memory progress through states with each state representing the values of the elements of the representation scheme. Capturing these different states can allow the user to "walk back" from the resulting solution to see how the solution was derived.

The evaluation procedure V assesses the result of each stage of the learning process, e.g., the transformation of data as represented in memory. This evaluation can be performed internally or externally. An example of internal evaluation in kmeans is the set of clusters resulting from one iteration of the algorithm. A typical evaluation is what percentage of the the elements in each cluster have moved from another cluster. A typical stopping criterion is when no points have moved from the previous stage. An external evaluation could be the user entering a command to terminate or a value indicating how P did.

Stopping critieria are often specified externally in order to avoid systems "running way" through oscillation over results that are "almost correct, but not good enough yet" or that continue generate representations that never meet a threshold for algorithm or reasoning termination.

I suggest that that this is a good way to start thinking about your project, whether you choose an MLS approach or other type, for your project. If you have questions or need assistance working through this, I will be at Foundation about 8:30 AM each class day, and can probably stay after the class ends at 11:30 AM.

Paul Scott (Scott 1983) developed a Taxonomy of Machine Learning Systems (MLS).

According to Scott, a learning system, LS, is a system which changes its behavior in response to modifications to its memory. It operates within an environment and consists of several components. Essential to this definition is that a learning system has a memory which evolves over time. The memory is comprised of a set of representations that organize the knowledge gained from its experiences. The learning system must be able to incorporate experiences based on its interpretation of events, performance of tasks, and their subsequent evaluation. Otherwise, it functions as a purely deterministic automaton. Additionally, it may or may not interact with its environment to assess its performance based on the completion of tasks.

      Scott identified both structural and functional components of an MLS.  The structural components are:

•       RS              a representation scheme
•       E                a set of possible experiences
•       Q               a set of values of representations
•       M               a set of internal memory states

      The functional components are:

•       P                a learning procedure
•       G               an experience or problem generator
•       V               a representation evaluator or critic

            So, a learning system can be described by a 4-tuple:

     LS = {S, P, G, V}

where S is the structural description of the learning system: S = { RS, E, Q, M }.

     With is description in mind, we can describe an eightfold model of learning systems:

The eight types can be organized along three functional components. The division along P leads to two groups:

• Self-organizing Group:                       (PVG, PG, PV, P)
• Teachable Group:                               (G, V, VG, Rote)

The eight types of learning systems are (with names assigned by Scott):

• PVG                Discovery Systems
• PG                  Reinforcement Systems
• PV                  Conceptual Clustering
• P                    Classical Conditioning
• VG                 Skeptical Learning
• V                    Advice-taking Systems
• G                    Inquiry Systems
• ()                    Passive Systems (learning by being told)

For the last category, we use the term rote learning.

An event generator G is a procedure that creates a set of events for input to the learning procedure of the learning system. Let us assume that there is a generator G which creates events e_j for LS. G may be internal or external to LS. The event generator may be a human (teacher), a program, or an automated sensor or device.

In the case of the human, he may not know the intended objective of the learning activity. So, it is possible for him to produce examples which do not contribute to the learning process. The event generator – be it human or program – may be distinct and independent from the critic who evaluates the results.

The four types for which G is an internal component are labeled G, VG, PG, and PVG in the above model.

G has the general form:

G: W x R_i* x M -> E x M’ (for all i)

where  W may be supervised or unsupervised

•             M, M’ represent memory state sets
•             E is the event history
•             R_i* is a representation history

Each instance of an execution of G takes the current representation and the current memory, interprets the event, and determines whether it must modify the representation and the memory.

A learning procedure P is a procedure that constructs or modifies the representations contained in memory. The changes to representations in memory are made through transitions of the representations. P has the general form:

P: E x M -> M’

Given a set of events E, P takes the memory M from its current state to a new state M’. Elements in M may have one or more representations or structures. Typically, P consists of a set of operators that transform elements

of M, e.g., p_i(e_j, m_k) -> m_k’.

An evaluation procedure V assesses the results of a transition and provides an evaluation. The evaluation may be performed internally by the learning system or externally by the world and provided to the learning system as an input. The evaluation is used to determine whether to accept or reject the transition. V may be loosely viewed as determining whether a particular change is “good” or “bad”. However, such evaluations need to be made in a larger context which must be explicitly specified in any description of V.

V has the general form:

V: E x M -> Z

where Z is a set of possible evaluations that V is able to make.

V can be simple or sophisticated. Some possibilities are:

1. Z may be some binary measure equating to “good” or “bad”
2. Z may be a numerical value – selected from a discrete or continuous interval
3. Z may be a complex structure that is used in making the transition

So, there are different types of learning systems. Think about this brief definition. Under what submodel would statistic ML fall?

I am Steve Kaisler, Adjunct Professor in the Department of Computer Science, School of Engineering and Computer Science.

I was born in Baltimore City, MD many years ago. I have lived in Maryland all my life - most recently for the past 36 years in Howard County, MD

I graduated from University of Maryland, College Park, MD with a B.S. in Physics (Minor: Computer Science) and an M.S. in Computer Science. I have a D.Sc. from George Washington University in Computer Science.

I have taught part-time at GWU - first in EECS, then in CS for over 42 years. I am interested in almost aspect of Computer Science as evidenced by the various topics of technical papers, tutorials, and published books - but more recently in Big Data and Analytics and Historical Computing Machines (six books). mY research interests included natural language processing, geopolitical simulation, and intelligent access to databases.

I am currently Primary Co-Chair of the Big Data and Analytics Minitrack at the Hawai'i International Conference on Systems Science and formerly Co-Chair of the Enterprise Architecture Minitrack.

I previously worked for the US Government in the defense community, as a Program Manager in Strategic Computing at the Defense Advanced Research Projects Agency (DARPA) and as Director of Systems Architecture and Technical Advisor to the Sergeant at Arms of the U.S. Senate. I also worked as Chief Scientist and as a Senior Research Scientist for several small businesses.

I have two children - both grown in their 30s - one boy and one girl, who has a daughter of 3-1/2 years old. I also have multiple cats who rule the house. I am an avid reader and like to travel and visit museums art galleries, and especially my daughter, her husband, and granddaughter in Canterbury, England.

Welcome to your brand new blog at GW Blogs.

To get started, simply log in, edit or delete this post and check out all the other options available to you.

For assistance, visit our comprehensive support site, check out our Edublogs User Guide guide or stop by The Edublogs Forums to chat with other edubloggers.

You can also subscribe to our brilliant free publication, The Edublogger, which is jammed with helpful tips, ideas and more.