Scribd Logo
Upload

Welcome to Scribd!

  • 486 views

    Artificial Inteligence NTU Tutorial 1 Answer

    Uploaded by

    bluepunchs
    The document discusses two different scenarios: 1. An agent searching a news website to find unread stories based on user-provided keywords. The agent must download pages, search texts, build a ranked story list, and handle errors, in an environment that can be dynamic as the website changes. 2. An agent in a maze with walls, trying to reach an alcove "A". Different sensor/memory configurations are discussed - no memory leads to failure, while 1-state memory allows the agent to succeed from any starting point.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Artificial Inteligence NTU Tutorial 1 Answer

    Uploaded by

    bluepunchs
    486 views1 page
    The document discusses two different scenarios: 1. An agent searching a news website to find unread stories based on user-provided keywords. The agent must download pages, search texts, build a ranked story list, and handle errors, in an environment that can be dynamic as the website changes. 2. An agent in a maze with walls, trying to reach an alcove "A". Different sensor/memory configurations are discussed - no memory leads to failure, while 1-state memory allows the agent to succeed from any starting point.

    Original Description:

    Tutorial answer for the first tutorial of CZ3005 (Artificial Inteligence) course in NTU, Singapore

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses two different scenarios: 1. An agent searching a news website to find unread stories based on user-provided keywords. The agent must download pages, search texts, build a ranked story list, and handle errors, in an environment that can be dynamic as the website changes. 2. An agent in a maze with walls, trying to reach an alcove "A". Different sensor/memory configurations are discussed - no memory leads to failure, while 1-state memory allows the agent to succeed from any starting point.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    486 views1 page

    Artificial Inteligence NTU Tutorial 1 Answer

    Uploaded by

    bluepunchs
    The document discusses two different scenarios: 1. An agent searching a news website to find unread stories based on user-provided keywords. The agent must download pages, search texts, build a ranked story list, and handle errors, in an environment that can be dynamic as the website changes. 2. An agent in a maze with walls, trying to reach an alcove "A". Different sensor/memory configurations are discussed - no memory leads to failure, while 1-state memory allows the agent to succeed from any starting point.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 1

    1 CZ3005 Tutorial 1

    1. a. Goals : find all unread stories that contain the specified keywords (periodic search)
    Performance measure : number of stories found? not news site entirely explored
    Environment : computer (agent), network, Internet (incl. the news site server itself), user.
    Actions : - send a download request to retrieve a page
    - search the text of a received page
    - build a ranked list of stories (links), display
    - handle error message etc.
    - query the user (keywords, stories read)
    Percept : - downloaded pages
    - messages from the agents computer (the news sites server, others)
    - user keywords
    - unread stories

    b. Observable : Fully if assuming no hidden / inaccessible page. It is partially otherwise.
    Deterministic : Not. Unexpected events e.g, dead links. Assume the news site is error-free? No, still
    communication may break down / time out. Assume a perfectly reliable internet? No,
    because there is no perfectly reliable internet
    Episodic : No. Sequential search, periodic but the agent must remember all stories read before.
    Dynamic : Yes. News site may change during search (page/stories added, edited, removed, etc.)
    Assume it never changes? not
    Discrete : yes, digital data and communication
    Suitable type of agent : goal-based agent. The task is essentially a search in a state space consisting of
    web pages
    Characteristic of interest : contingency problem. (agent must interleave search and execution actions
    can only be decided after downloading a page and searching its content)

    2. a. No memory, front sensor:
    IF wall ahead THEN turn left
    IF no wall ahead THEN move forward
    Agent initially facing A: trivial success
    Any other initial config: agent keeps following the wall anti-clockwise and never detects the alcove A

    b. No memory, front and right sensors:
    IF wall ahead AND right wall THEN turn left
    IF wall ahead AND no right wall THEN turn left
    IF no wall ahead AND right wall THEN move forward
    IF no wall ahead AND no right wall THEN turn right
    Agent initially has a wall ahead or on the right: success
    Agent initially not close to any wall: keeps turning right

    c. 1-state memory, front and right sensors:
    sensor: current state wall ahead, right wall
    memory: previous state wall ahead before, right wall before

    IF wall ahead THEN turn left
    IF no wall ahead AND // alcove detected
    ( no right wall AND right wall before ) THEN turn right
    IF no wall ahead AND not ( no right wall AND right wall before ) // no alcove
    THEN move forward
    Agent reaches the alcoves in all cases success!

    You might also like