A Genuine Learning Building Application: The NeuroScholar Venture

0
0
1897 days ago, 601 views
PowerPoint PPT Presentation
A Real-World Knowledge Engineering Application: The NeuroScholar Project. Gorge APC Burns ... survey article is a case of a non-computational information model ...

Presentation Transcript

Slide 1

A Real-World Knowledge Engineering Application: The NeuroScholar Project Gully APC Burns K. M. Look into Group University of Southern California

Slide 2

Structure of the presentation Ideas & Concepts Design Implementation Demonstration

Slide 3

I. Thoughts & Concepts In which we are helped to remember what a great many people think information is, the way it is as of now utilized (and abused) and how we may enhance matters.

Slide 4

What does "Learning" mean? Fundamental Entry: knowl·edge Pronunciation: 'nä-lij Function: thing Etymology: Middle English knowlege, from knowlechen to recognize, sporadic from knowen Date: fourteenth century 1 out of date : COGNIZANCE 2 a (1) : the reality or state of knowing something with commonality increased through experience or affiliation (2) : associate with or comprehension of a science, craftsmanship, or procedure b (1) : the reality or state of monitoring something (2) : the scope of one's data or comprehension <answered to the best of my insight > c : the situation or state of capturing truth or actuality through thinking : COGNITION d : the reality or state of having data or of being educated <a man of uncommon information > 3 obsolete : SEXUAL INTERCOURSE 4 a : the whole of what is known : the collection of truth, data, and standards obtained by humankind b antiquated : a branch of learning [from http://www.m-w.com/]

Slide 5

The distributed writing … is the finished result of research and in that capacity shapes the reason for human comprehension of the subject … is extremely significant. … is organized. … is interpretable. Picture taken from U.S. Topographical Survey Energy Resource Surveys Program

Slide 6

The distributed writing … is extensive and awkward. … has shifting dependability. … is conflicting. … depends on common dialect. … is hard to mechanize. … is pithy … is subjective … is 2-D Image taken from U.S. Land Survey Energy Resource Surveys Program

Slide 7

The distributed writing … is a legitimate focus for assault with informatics-based strategies. This licenses … (an) Increased illumination through formalization (b) substantial scale information taking care of capacity (c) investigation of existing information to inspect association Image taken from U.S. Geographical Survey Energy Resource Surveys Program

Slide 8

The present status of "hypothesis" in Neuroscience How we might want neuroscientists to think Where we might want to work A semantic continuum [Mike Uschold, Boeing Corp] Shared human accord Semantics hardwired; utilized at runtime Semantics handled and utilized at runtime Text depictions Implicit Informal (unequivocal) Formal (for people) Formal (for machines) Further to the right means: Less uncertainty More prone to have remedy usefulness Better between operation (ideally) Less hardwiring More hearty to change More troublesome

Slide 9

What's the matter with this photo? … from a neuroscientist's perspective… Number of structures = 500 x 2 Number of Cell Groups per structure = 10 Number of Possible Connections between cell bunches = 10,000 x 10,000 = 10 8 Estimated Number of Connections between cell bunches = 250,000 From Swanson (1998), "Cerebrum Maps, Structure of the Rat Brain", 2 nd version, Elsevier, Amsterdam.

Slide 10

… it's far and away more terrible than that … Neuroscience is to a great degree multidisciplinary Spatial Scales of Measurement: 10 1 – 10 - 9 m Temporal Scales of Measurement: 70 yrs (2.21 x 10 9 s) to 10 - 3 s (not notwithstanding including transformative time!) Study happens in a heterogeneous hypothetical system including: Anatomy, Physiology, Psychology, Ethology, Biochemistry (Molecular Biology, Genetics, Bioinformatics), Biophysics, Behavioral Ecology, Biology … to give some examples… All of these subjects are specific, difficult to connection work amongst controls and crosswise over levels

Slide 11

… & it's surprisingly more terrible than that !!! Neuroanatomical classification are the nearest thing that neuroscience has for an institutionalized system… In any given paper, a similar name might be utilized for various structures, or distinctive names might be utilized diverse structures. e.g., 'Globus Pallidus, standards medialis (GPm)' additionally called the 'Entopeduncular Nucleus' by others. See the file of Swanson (1998), "Mind Maps, Structure of the Rat Brain", 2 nd version, Elsevier, Amsterdam rundown of equivalent words as per one source.

Slide 12

We confine the issue space to a particular dissolvable system Describe a given wonder (e.g., the anxiety reaction). Distinguish which populaces of neurons are included in the marvel (i.e., any neurons that turn on, kill, change their terminating, influence the wonder if upset, and so on ). Speak to how these populaces of neurons are interconnected. Speak to the dynamic procedures of there neurons that underlie the wonder.

Slide 13

A Construct: 'A Knowledge Model' = A customized representation of an individual's learning . e.g., A survey article is a case of a non-computational information display

Slide 14

Another Construct: 'Learning Landscape' = A guide of Knowledge Models ( where every KM is timestamped ) e.g., A rundown of the best audits of a given subject after some time is a case of a non-computational learning scene

Slide 15

II. Plan In which these vain thoughts are put into a coherent outline and it turns out to be obvious that the outline criteria of the NeuroScholar extend recognize it from unadulterated research in software engineering

Slide 16

Some outline necessities all together of significance Powerful & empowering to neuroscientists in their ordinary work Easy to utilize! ( i.e., free, multi-stage, a single tick establishment) Knowledge obtaining/information examination is the rate restricting stride Open-hotspot for future advancement as a scholarly venture.

Slide 17

Knowledge Landscapes NeuroScholar Screenshot-(sham information)

Slide 18

'Learning Landscape' 'Information Collection' "Pieces" 'Learning Model' "Substances" "Properties" "Relations" "Explanations" Knowledge Landscapes NeuroScholar Screenshot-(sham information)

Slide 19

'Information Collection' An arrangement of information sections "Comments" Knowledge Models & cases e.g. a production: Allen GV & DF Cechetto. (1993) J Comp Neurol 330:421-438. "Sections" "Substances" "Properties" "Relations"

Slide 20

singular bits of the writing "Pieces" "Comments" Knowledge Models & illustrations 'Information Collection' e.g. depictions of exploratory results. "… Moderate to light terminal naming was available in the parvocellular bits of the paraventricular core, front hypothalamic core, foremost segment of the sidelong hypothalamic territory (Figs. 2D, 3B), and in the focal core of the amygdala (Fig, 2D)… ." From Allen & Cechetto (1993) "Elements" "Properties" "Relations"

Slide 21

e.g. neuronPopulation protest learning sort = portrayal space sort = tract-following analysis "Substances" brainVolumes experimentalMethod naming "Properties" "Explanations" injectionSite marking Knowledge Models & cases Abstract information structures that catch the significance of an arrangement of pieces inside the structure of the NeuroScholar framework 'Information Collection' "Sections" "Relations"

Slide 22

ZI LHA "Comments" Knowledge Models & illustrations Rules that connection two questions together. 'Information Collection' "Parts" "Substances" "Properties" "Relations" "Relations"

Slide 23

Sets of articles and relations, unequivocally chose and organized inside framework neuronPopulation2 "Explanations" neuronPopulation1 Knowledge Models & cases 'Information Collection' "Sections" "Synopses" "Elements" "Properties" "Relations"

Slide 24

"Comments" Human-interpretable content to make substance of learning base justifiable "Comments" Knowledge Models & cases 'Information Collection' "Pieces" "Objects" "Properties" "Relations"

Slide 25

Distributed Online Sources of Information "Parts" Local Implementation

Slide 26

Distributed Online Sources of Information Users' Spaces & Models "Parts" Centralized Published Knowledge Repository Local Implementation

Slide 27

Distributed Online Sources of Information "Pieces" Users' Spaces & Models 'Pending Review'

Slide 28

Distributed Online Sources of Information "Sections" Users' Spaces & Models P2P sharing Knowledge Model Comparison

Slide 29

Knowledge Model Comparison Given two clients A & B, with Knowledge Models K A & K B being shared under the P2P display. We need A to have the capacity to run a program that consequently thinks about K B to K A so that the disparities and inconsistencies between the two models can be comprehended and accommodated.

Slide 30

What's the matter with this photo? … from a PC researcher's perspective… Where is the formal rationale? It's o.k. on the off chance that we just fare learning models to a formal rationale based representation rather that construct our whole approach in light of it. Information Acquisition is the rate-constraining stride!

Slide 31

Knowledge Representation Knowledge representation is a multidisciplinary subject that applies hypotheses and methods from three different fields: Logic gives the formal structure and principles of induction. Philosophy characterizes the sorts of things that exist in the application space. Calculation underpins the applications that recognize learning representation from immaculate theory… Sowa (2000), Knowledge Representation: Logical, Philosophical, and Computational Foundations , Brooks Cole Publishing Co., Pacific Grove, CA.

Slide 32

Knowledge Representation … Without rationale, a learning representation is

SPONSORS