2543 days ago, 878 views
PowerPoint PPT Presentation
Paul Dunne GMIT. Computerized reasoning. 2. AI recommends that canny conduct can be accomplished through the control of image structuresYou could utilize this information to derive new facts.Galway has another flame engine.It must be red!We speaks to actualities (on the PC) utilizing information structures and we compose project code to prevail upon them..

Presentation Transcript

Slide 1

Presentation What is learning? Attempt to characterize it. The investigation of learning is a called epistemology. Information can be further characterized as Procedural learning Knowing how to accomplish something (eg knowing how to begin an auto.) Declarative information Knowing that something is valid or false Tacit information eg I know how to put the key in the start yet do I know how my hand/arm/muscles/ligaments/nerves all corordinated? Computerized reasoning

Slide 2

Fire-Engine Symbol Structure Symbols Red (Fire-Engine) Red Introduction AI recommends that keen conduct can be accomplished through the control of image structures You could utilize this information to surmise new actualities. Galway has another fire-motor. It must be red! We speaks to realities (on the PC) utilizing information structures and we compose program code to prevail upon them. Manmade brainpower

Slide 3

Knowledge Representation Languages High level information portrayal dialects have been created to make learning portrayal less demanding We'll take a gander at a couple of these Semantic nets Frames Predicate rationale Rules Artificial Intelligence

Slide 4

First however… Requirements for a learning portrayal dialect It must permit you speak to satisfactorily complex certainties in a reasonable and exact yet common way , and in a way that effectively permits you to find new truths from your current learning. satisfactorily mind boggling actualities Known as - Representational sufficiency A level of detail that will deliver the outcomes we require and no more if a straightforward portrayal dialect will be satisfactory then utilize a basic dialect! detests (everyone, grows) Artificial Intelligence

Slide 5

Requirements for Knowledge Representation Languages clear and exact Known as - Well characterized linguistic structure and semantics no space for perplexity and no ambiguities. Linguistic structure characterizes the reasonable structures which depict how to make sentences "Aversions – everyone – sprouts" is not permitted Semantics characterizes what it signifies "despises (everyone, grows)" implies everyone detests grows not grows hate everyone . Characteristic Language must not be excessively mind boggling and hard to decipher. conclude new realities Known as - Inferential ampleness The dialect must bolster deduction. Manmade brainpower

Slide 6

The principle contenders Broadly talking there are three fundamental ways to deal with learning portrayal in AI. SEMANTIC NETS/FRAMES Easy to utilize and generally common. Rationale Arguably the most imperative. With an all around characterized language structure and semantics. Permits surmising yet can be hard to make an interpretation of this present reality into rationale. Experiences issues with time, vulnerability and convictions. RULES (If fire Then yell help) Condition-activity tenets or Production Rules(specifying what to do under specific conditions) inside a lead based framework. Computerized reasoning

Slide 7

Semantic Nets/Frames Artificial Intelligence

Slide 8

Semantic Networks A basic class chain of command Allows you to speak to classes (or classifications) of items and relationship amongst articles and to draw straightforward surmisings in light of this information. The net is comprised of hubs and circular segments/joins interfacing the hubs. The connections express connections. Without connections information is essentially an accumulation of inconsequential certainties. – with them other learning can be derived. (nellie has a head!) The hubs express physical articles, ideas or circumstances. Close: Clyde and Nellie both have heads Artificial Intelligence

Slide 9

Semantic Networks Originally used to speak to the importance of English words The connections speak to the connections . The most essential relations are; known as an A-KIND-OF (AKO) relationship known as a Seems to be A relationship Other relations likewise permitted To speak to properties of articles and classifications of items. Semantic systems ordinarily permit proficient legacy based derivations utilizing extraordinary reason calculations. Semantic nets in maths are marked, coordinated charts. Some of the time known as cooperative nets (i.e. hubs are related or identified with different hubs) Artificial Intelligence

Slide 10

Semantic Networks Have a Go ! - Class Problems Represent each of the accompanying helpful bits of learning as a semantic net. (a) "Floyd is a little hippo who lives in Dublin zoo. Like all hippos he eats grass and likes swimming" (b) "The aorta is a specific sort of supply route which has a width of 2.5cm. A conduit is a sort of vein. A conduit dependably has a strong divider, and by and large has a measurement of 0.4cm. A vein is a sort of vein, however has a sinewy divider. Veins all have tubular frame and contain blood." Artificial Intelligence

Slide 11

(a) (b) Artificial Intelligence

Slide 12

Semantic Networks To attempt and legitimately characterize the semantics of a semantic system (what it implies) set hypothesis is frequently utilized. Semantic systems permits us to speak to learning about items and connections between articles in a natural way.  However the kind of deductions that can be bolstered is genuinely prohibitive (only legacy of properties).  Also the absence of any gauges for connection names is risky.  A semantic net can't characterize learning and in addition rationale can.  There is no real way to encode heuristic learning (dependable guidelines) into semantic nets. It remains a decent decision as a learning portrayal dialect for some AI issues – especially to show twofold connections. Counterfeit consciousness

Slide 13

Frames are a variation of semantic systems prominent approach to speak to truths in a specialist framework. The distinction is fundamentally in the level of detail of a hub. In semantic nets the hub has a name . Properties of a hub are indicated utilizing different hubs and some kind of relationship connecting them. In an edge the hub itself has a structure. This implies the hub can contain values or to be sure different edges. Semantic nets (by and large) speak to learning about a wide zone. Outlines speak to (related) learning about a tight subject A casing would be a decent decision for portraying something, for example, an auto – a PC and so forth. Computerized reasoning

Slide 14

A space esteem may likewise be an edge Property and Slot phrasing is tradable Frames An edge is fundamentally a gathering of openings (once in a while called properties) and opening qualities or fillers) that characterize a cliché protest. Counterfeit consciousness

Slide 15

Frames Some of the wording from edges has been embraced by Oject Orientated innovation. It is straight forward to decipher between semantic systems and edge based representaion. Class and Instance Nodes - Objects Links - Slots Node at end of Link - Slot esteem Artificial Intelligence

Slide 16

Semantic Net and Frame Artificial Intelligence

Slide 17

Defaults and Multiple Inheritance Objects acquire the default property estimation unless they have an individual property estimation that contentions with the acquired one. Different legacy is troublesome and regularly brings about clashes. Nellie could be a bazaar creature who has been uncommonly "designed" to be white. Nellie then is a subclass of Elephant and Circus-creature. Shading? The incentive for Elephant or for Circus-creature? The framework must accommodate these contentions – with the end goal that it will give back the correct an incentive for shading (white) and the correct an incentive for size (vast). Manmade brainpower

Slide 18

Practical Frames The openings and space qualities can be casings. The space qualities could likewise be methods, executed when the incentive for the opening is required. [The framework is then depicted as having a procedural rather that decalarative semantics] Implementation of a straightforward casing framework could be completed with the accompanying calculation [Note the recursion]; What is the opening an incentive for protest "O's" space P. value(O, P) If opening name P exits for protest O then Return space estimation of P Else if question O has space called "subclass" or "occasion". At that point return "Super", the estimation of this opening Find value(Super, P) and give back this esteem. Else Fail. Computerized reasoning

Slide 19

Frames and Semantic Nets Summary Frames and semantic systems give an unmistakable and straightforward method for speaking to properties of items and classifications of articles. A fundamental deduction is accessible through legacy. It doesn't adapt will to Negation ( not A) Disjunction (An or B) Quantification (for each of the An and some B) Artificial Intelligence

Slide 20

Frames Go on attempt it. Speak to the accompanying as edges; (a) "Hippos live in Africa. Hippos are by and large very expansive. Floyd is a little hippo who lives in Dublin zoo. Like all hippos he eats grass and likes swimming" (b) "The aorta is a specific sort of vein which has a distance across of 2.5cm. A supply route is a sort of vein. A vein dependably has a strong divider, and for the most part has a measurement of 0.4cm. A vein is a sort of vein, yet has a stringy divider. Veins all have tubular shape and contain blood." Artificial Intelligence

Slide 21

Frames (b) Blood-vessel frame tubular contains blood (a) Hippo likes swimming eats grass *lives Africa *size huge Artery ako Blood-vessel divider solid *Diameter 0.4cm Aorta is-an Artery Diameter 2.5cm Vein Ako Blood-vessel Wall fiborous Floyd is-a Hippo lives Dublin Zoo measure little Artificial Intelligence

Slide 22

Logic Artificial Intelligence

Slide 23

Logic Most critical information portrayal dialect is seemingly (predicate) rationale. It permits us to speak to genuinely complex certainties about the world, To infer new actualities in a manner that ensures that if the underlying realities were genuine then so are the conclusions. The expression "thinking", "derivation" and "reasoning" are by and large used to cover any procedure by which conclusions are come to. Rationale is a formal framework which might be portrayed as far as its sentence structure , sematics and its evidence hypothesis . To begin with we will take a gander at the less difficult propositional rationale (infrequently c