Course: Artificial intelligence in applications. Modeling, Machine Learning and Data Classifier Performance | moodle-thab

  • General

    • Wednesday same time (8:00-9:30 & 9:45 - 11:45) Lectrures & excercises in person, Building 20, room 211 Forum
      Students must
      Mark as done

      Dear students,

      On Wednesday, 27.11.2024, and all following lectures we will continue with the lectures and excercises in our Artificial Intelligence courses as usual (not any online meetings).

      We meet as plant in  building 20 - room 211

      8:00 - 9:30 Uhr
      V Cognitive and object-oriented modeling - under uncertainties in knowledge and data - as aspects of a
      Rechnerraum 20-211

      and

      9:45 - 11:15 Uhr
      V Artificial intelligence in applications. Modeling, Machine Learning and Data Classifier Performance
      Rechnerraum 20-211

      See you on Wednesday, in C1/20-211 !

      Prof. Dr. Galia Weidl


    • List of topics to select for your Student's Presentation on Bayesian Networks. Selecting completely free your own topic of interest is encouraged! File
      Students must
      Mark as done

      Plan for the students' presentations in WiSe2024/25: 

      You can select a topic from the list of available topics/materials or you can search for new materials and/or you can prepare your own example.

      The dates are tentative for the presentation in the class. 

      We can decide to shift them all by 1-2 weeks later. Everyone should have one presentation in the class.

      First, you do the  presentation (20 min) and then we model (15-20 min) together with everyone in the class the model you have presented.

      The oral exams for everyone will be probably on 21/22.01.2025, after we get an official booking from the dean office.

      Prof. Galia Weidl
      ___
      If you prefer to prepare your own application: How can this presentation be arranged?
      If you chose a topic (e.g. a small real application) where you need to prepare the (e.g. found from internet) data for learning of the probabilities of your model, so you can split the work into two:

      - one student can work on the structure of the model and collect all information necessary to build the structure ( graph of the network)
      - second student will then learn from data the probabilities for the created structure.

      In the class I will show you with one automotive example (maneuver recognition for cars) how this can be done. And then you can do it in a similar way.

      Where to find data?
      My PhD students will show you (in the comming classes, im November/December) some web-pages, where you can find data to build your application.

    • _Course 7540: Artificial intelligence in applications. Modeling, Machine Learning and Data Classifier Performance Lesson
      Students must
      Mark as done

      Schedule every Wednesday, 9:45 - 11:15

      -------------------------------------------------------------------------------------
      General Information: each course has 2 ECTS      , according the the following structure:

      7525: Cognitive and Object-Oriented Modeling Under Uncertainties as Aspects of Artificial Intelligence in Practical Applications. (=2 ECTs)

      7540: Artificial intelligence in applications. Modeling, Machine Learning and Data Classifier Performance.(=2 ECTs)

      Your personal contribution will be organized as follows:
      • choose a topic/problem you would like to model with Bayesian Networks
      • do the correspondiing study of related publications on the topic of your choice (this will be the background for the development of your model),
      • Modelling: choose a model from the literature to present OR develop your own model, according to the modeling princiles we have learned in the course 7525 (= 2 ECTs)
      • search for open source data on your topic in order to learn the parameters of the model   (Maschine Learning as we will learn in course 7540) = 2 ECTS
      • For the final exam: 20 min  presentation of a model of your choice and answer questions (listed in moodle) - for each course

      Only for the incoming students: 

      In case you prefer to participate only in 7525 or only in 7540: 

      In each course 7525 or 7540: give 20 min presentation on Modelling and 15 min modeling uin the class together, so that everyone can reproduce the model you developed.

      For  participants in both courses, you will have in each course: 

      course 7525: 20 min presentation on Modelling and 15 min modeling together, so that everyone can reproduce the model you developed.

      course 7540: 20 min on Learning with Data for the model you have developed 

      The exam for eah course will consist of 15 min presentation on the selected subject and covers the questions which will be given in the class.
      --------------------------------------------------------------------------------------------------------------------------------------------
      Note for the incoming students:
      International Office – Incoming Coordinator: Carina Blaeser
      Aschaffenburg University of Applied Sciences
      ----------------------------------------
      I wish everyone a lot of success!
      Prof. Dr. Galia Weidl


    • Process on the Student Research Project: BN for Road Congestions File
    • Exam Planning for both courses with Prof. Galia Weidl in January 2025 (... Artificial Intelligence) Forum
      Please book your time slot for the exam in January 2025 here:
      https://moodle.th-ab.de/course/view.php?id=7293
      ---------------------------------------------------------------------------------------
      To allow booking, you need first to subscribe to the moodle page  

      Exam Planning (Galia Weidl)

      Einschreibeschlüssel: AI_MyChoice1

      ---------------------------------------------------------------------------------------
      This is the exam planning  (oral presentation)  for both courses:

      Course 7525: Cognitive and object-oriented modeling under uncertainties as aspects of artificial intelligence in practical applications
      &
      Course 7540: Artificial intelligence in applications. Modeling, Machine Learning and Data Classifier Performance

      Please select in (Prüfung) Planer one time slot  (Zeitfenster) of 15min if you are taking just one course
      or 2 time slots together 30 min (in case you have 2 x 15 min) for your oral presentation 

      Tuesday AFTERNOON (21.01.2025 in Campus C3/Room 107, 

      Exam time 14:00 - 19:00, 

      OR on 

      Wednesday (22.01.2025), 08:00 - 18:00, in Campus C1/Building 04/Room 203 

      Exam time 08:00 - 17:00

      I wish you a lot of success with your exam!

      Danke & Viele Grüße / Many thanks & Best regards,

      Galia Weidl  

      Prof. Dr. Galia Weidl


    • Mentorat (Consultation/Sprechstunden) Page
      Students must
      Mark as done
    • The lecture & excercises on Wednesday, November 6. & 13. + December 4th 2024 Forum
      Students must
      Mark as done

      Dear Students,

      The lecture & excercises on Wednesday, December, 4th. will be given by my PhD Students.

      If you have questions about where to get some data for your model and how to prepare the data for learning the parameters of your model, you can ask them to help you.

      Best regards,

      Galia Weidl


  • Open all

    Close all

  • Instructions: Clicking on the section name will show / hide the section.

  • ´2024-10-23 – 2024-11-20  – Prof. Dr. Galia Weidl – Learning in Bayesian Networks, using the Hugin tool, Applications of Bayesian Networks

    Basics on Data sources, - preprocessing, Techniques and Tools for learning in OOBN Applications

       ´20/11/2024  & 04/12/2024: Mofeed Chaar – data processing with python   
       ´27/11/2024 & 04/12/2024 -– Kranthi Talluri – case study on traffic congestion – from data (processing) to application (cognitive modeling and parameterization)  results: recognition of congestion

    ´2024-10-23: Dr. Jamal Raiyn, Artificial Intelligence, Introduction
    ´2004-12-11: Dr. Jamal Raiyn, Machine Learning, and WEKA basics

    ´2024-12-18: Dr. Jamal Raiyn, Machine Learning, and WEKA excercises with practical Data Mining

    ´Distribution of Students’ Presentations
          ´WiSe2024 – Students’ presentations of selected applications
          ´Fix the dates for Students’ presentations of selected applications
    ´Exams:
    2025-01-22 – first day of exams



  • This section

    2024-10-23: Introduction. Machine Learning - Lecture 1 by Dr. Jamal Raiyn
    • Bayesian Network for Analysis and Prediction of Traffic Congestion Using the Accident Data File
      Students must
      Mark as done

      This topic "Bayesian Network for Analysis and Prediction of Traffic Congestion Using the Accident Data" has been also presented by a PhD Student Kranthi Talluri at the 10th international conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2024):

      https://www.insticc.org/node/technicalprogram/vehits/2024/

      The zip file above contains all relevant materials, data, preprocessing for modeling and the final BN model.

      If you are interested, you can wokr with this data to prepare your own presentation. Alternatively you can use one of the open source data sets. Kranthi will be happy to help you with the preparatioon of data and BN model .

  • The original presentation was given on 

    2023-11-20 Student's Presentation: Object Oriented Bayesian Networks and Application (Accident Cars) by Johannes Krausert (s201020)

    You can use all these materials in this section to prepare your own presentation. 

    https://download.hugin.com/webdocs/manuals/9.4/htmlhelp/pages/Tutorials/ObjectOrientedBayesianNetwork/IntroductionToObjectOrientedNetworks.html

    https://download.hugin.com/webdocs/manuals/9.4/htmlhelp/pages/Tutorials/ObjectOrientedBayesianNetwork/HowToBuildAnObjectOrientedBayesianNetwork.html

    https://download.hugin.com/webdocs/manuals/9.4/htmlhelp/pages/Manual/AboutBayesianNetworks/ObjectOrientedNetwork.html

  • Based on a lecture from 2022-12-08: Lecture: How to read the paper on idioms and chapter 6 of the eBook? Continuation from 2022-11-17 Bayesian Networks for Decision Making in Automotive Applications 

  • The lecture on December 8th will be on-line and as Exercices on your own. 

  • Laphutama Wutthichai

  • Simon Hain, 

    You can adress your questions to Hain, Simon <s220155@th-ab.de>

  • Presentation on 2023-12-11 (18?): 

    Donghui Li s230292

    Melike Damla Yeter s230365

    Anda Dong s230366

    Learning in BN

    https://download.hugin.com/webdocs/manuals/9.4/htmlhelp/pages/Tutorials/Structure_EM_Adaptation_Learning/StructureLearningTutorial.html

    • EM Learning of parameters for the BN model structure

    https://download.hugin.com/webdocs/manuals/9.4/htmlhelp/pages/Tutorials/Structure_EM_Adaptation_Learning/EM_LearningTutorial.html

    • Adaptation: 

    https://download.hugin.com/webdocs/manuals/9.4/htmlhelp/pages/Tutorials/Structure_EM_Adaptation_Learning/EM_LearningTutorial.html

  • ---------------------  Questions to cover in your presentation - in English  ------------------------------------------------------------------------------------------------------

    These questions  are  also relevant for the examination (mündliche Prüfung 20 Minuten)

    - Cognitive Hypotheses for Maneuver Recognition / or Cognitive Hypotheses for your own BN application

    • How is a BN (Bayesian Network) model build/generated/learned?
    • Knowledge based modeling (encoding causal relations in the BN structure) & Learning (of BN parameters from data) . The combination of knowledge and data leads to probabilistic decision making.
    •            What could be the source of knowledge and data?
    • What features are used in the hypotheses?  The features are obtained from measured sensor data or from computed data (also known as soft sensor data).
    • Why do we need to model sensor uncertainties for decision making? How? What is the influence of uncertainties on the decision?
    • Data used for Learning: Variables types (boolean, numbered, interval, labeled) 
    • What means "Data Labeling"? e.g. true/false, maneuver type (lane marking crossing: left/right/straight)?
    • Classification of hypotheses (e.g. driving maneuver) under uncertainties?
    • Evidence as input to the Bayes Net (Decision Making) Models and interpretation of classification (decision) results.

    • How to use a (software) tool (e.g. Hugin) for design, simulation and testing of decision-making model?

    • Split the data in training and test set. Show the Classifier Performance with the help of the Analysis Wizard, 

    • Show the results of the ROC with TP, FN, FP, FN or other metrics for the model evaluation.

    • When to use Bayesian Networks ? (discussed in the first lecture AND in the eBook, Chapter 1: Introduction; 1.5 When to Use Probabilistic Nets?) 

                                       eBook: Kjærulff, U. B., & Madsen, A. L. (2012). Bayesian networks and influence diagrams : A guide to construction and analysis


    ------------------------------------------  Gleiche Fragen für die Präsentation - auf Deutsch (Sie entscheiden auf welche Sprache wollen Sie die 20 min mündliche Präsentation bei der Prüfung machen --------------------------


    Wir diskutieren kurz welche Fragen sind Prüfungsrelevant. Unten habe ich einige Fragen aufgeschrieben. Die Studierenden:

    • designen (erzeugen/lernen) ein Entscheidungsfindungsmodell (Bayes Netz) von einer Domäne (an Beispiele)
    • führen ein Modell aus
    • kennen wissen- und datenbasierten Modellierung
    • kennen die Quellen von Wissen, Data und Unsicherheiten einer Domäne
    • kennen was Merkmale sind (die Charkteristische Eigenschaften/Parameter, welche ein Domain beschreiben). Die Merkmale  wurden erhalten aus gemessene Sensor Daten oder von berechnete daten (aus Modelen)  or  (auch bekannt als "soft sensor data").
    • Wie entstehen Hypothesen aus Merkmalen?
    • können Sensor-Unsicherheiten modellieren und analysieren. 
    • Warum ist es wichtig die (Sensor) Unsicherheiten auch bei Entscheidungsfindung zu berücksichtigen? bzw. Was ist der Einfluss von Unsicherheiten auf der Schlussfolgerung?
    • einordnen welche Merkmale werden in die Model-Hypothesen verwendet?  Die Merkmale werden erhalten von gemessene Sensor Daten  oder von berechnete Daten  (auch bekannt als "soft sensor data")
    • Welche Datentypen werden fürs Modelieren/Lernen verwendet: Variablen Typ (boolean, numbered, interval, labeled) 
    • Was ist "Data Labeling"? z.B. true/false, bzw. in Beispiel: Manöver Klass (Überquerung der Spurmarkierung: Links/Rechts/Gerade)?
    • Klassifizierung von Hypothesen (z.B. in einer Anwendungen wie Erkennuing von Fahrmanöver) unter Unsicherheiten
    • Evidenz als input in Bayes Net (Entscheidungsfindung) Model und Interpretation von Klassifikation (Entscheidung) Ergebnis.
    • kennen Werkzeuge für Design, Simulation und Testen von Entscheidungsfindungsmodelle

      • Split the data in training and test set. Show the Classifier Performance with the help of the Analysis Wizard, 

      • Show the results of the ROC with TP, FN, FP, FN or other metrics for the model evaluation.


    • einordnen für welche Anwendungen Entscheidungsfindung Methoden (wie Bayes Netze) angebracht und gut geeignet sind
             (diskutiert in der erste Vorlesung UND in eBook, Chapter 1: Introduction; 1.5 When to Use Probabilistic Nets?) 


  • ------------------------------------------ Relevant Exam Questions in English  ------------------------------------------------------------------------------------------------------

    - Cognitive Hypotheses  of your own BN application or an example of your choice (can be also from existing application you have selected to present)
    - Cognitive Hypotheses for Maneuver Recognition (as discussed in the lectures)

    We discuss what questions are relevant for the exam (oral examination 20 minutes for each course 7525 & 7540).

    The students can:

    • design (create/learn) a decision-making model (Bayesian network) from a domain (on examples of their own choice)
    • run a model
    • know knowledge- and data-based modelling

    o   encoding causal relations (in the BN structure) & Learning (of BN parameters from data) . The combination of knowledge and data leads o probabilistic decision making.

    • know the sources of knowledge, data and uncertainties of a selected domain
    • know what features are the characteristic properties/parameters that describe a domain.

    o   The features can be obtained from measured sensor data or from calculated data (from models), also known as "soft sensor data", or from communicated (mobile) data.

    • How do hypotheses arise from features/characteristics? - by combining the features/characteristics/variables that constitute/manifest a hypothesis.

          --> see slide 9 in the lecture  slides (in English): Bayesian Networks for Decision Making in Automotive Applications - see

                                                                                              https://moodle.th-ab.de/course/view.php?id=5539#section-18

    Lecture ( 2023-05-04 - ,...) Bayesian Networks for Cognitive & Object-orriented Modelling (Decision Making) in Automotive ApplicationsDatei 

           --> see Note below: Explanation for decision-making through hypotheses from characteristics 

    Note:

           * The Hypotheses are generated from the Features. In other words: from the Features are contructed the Hypothesen.

           * The hypotheses are evaluated, based on the Evidence (=measured Features) and this allows to make the decision.

           The Hypotheses are used in two different types of decision tasks:

    1. Diagnosis:

    For Diagnostics: die decision finds what Hypothesis has the highest probability, based on the observations/measurements (evidence). An example of Diagnostics task is the angina.oobn Bayes Network.

             2. Prognosis:  

    Different Hypotheses are combined in order to build the prognosis um eine Prognosis, e.g. for the Recognition of maneuvers, based on observed features.

    In the example " Recognition of maneuvers": The decision (=Maneuver Recognition) has several states: DONTCARE, OBJCUTIN, OBJCUTOUT, LANEFOLLOW, OBJFOLLOW, EGOOCUTIN, EGOCUTOUT. 
    These states are computed based on the evidence (= observed features).
    For example, the Lateral Evidence is a Hypothesis, which is build from the features (lateral velocity and laterale distance (offset).
    The features are measured and set/entered as evidence in the Hypothesis.
    A measured value (e.g. speed) is influenced by the really measured Variable (velocity) und by the uncertainty/inacuracy of measurement/Noise. 
        The decision finds what combination of Hypotheses (with the corresponding features) - in what state - has the highest probability. The ranking of states according to their probabilities allows to set priority on different states of the decision. 


    • can model and analyze sensor uncertainties.

    o   Why is it important to take the (sensor) uncertainties into account when making decisions? or What is the influence of uncertainties on the conclusion/decision?

    • understand which characteristics are used in the model hypotheses?

    o   The characteristics are obtained from measured sensor data or from calculated data (also known as "soft sensor data")

    • Which data types are used for modelling/learning:

    o   variable type (boolean, numbered, interval, labeled)

    • What is data labeling?

    o   annotate the state of a variable as e.g. boolean: true/false, or in example: maneuver class (variable “crossing the lane marking”: labeles: left/right/straight)

    • Classification of hypotheses under uncertainties?

    o   e.g. in an application such as detection of driving maneuvers

    o   based on the model structure and parameterization

    • Evidence as input to Bayesian network (decision making) model and interpretation of classification (decision) outcome
    • know tools for design, simulation and testing of decision-making models
    • understand for which applications decision-making methods (such as Bayesian networks) are appropriate and well suited (as discussed in the first lecture AND in eBook, Chapter 1: Introduction; 1.5 When to Use Probabilistic Nets)

     

    ------------- (Same content as above: in German///  Gleiche Fragen auf Deutsch, falls Sie Ihre Präsentation auf Deutsch vorbereiten oder präsentieren möchten) ------------------------------

    - Kognitive Hypothesen  unter Unsicherheiten in der gewählte BN Applikation (eigene oder gefundene Beispiele aus Internet/Literatur/etc.)
    - Kognitive Hypothesen  unter Unsicherheiten für Manöver Erkennung (siehe die Vorlesung)

    Wir diskutieren welche Fragen sind Prüfungsrelevant (mündliche Prüfung 20 Minuten für jede Kurse 7525 & 7540). Die Studierenden:

    • designen (erzeugen/lernen) ein Entscheidungsfindungsmodell (Bayes Netz) von einer Domäne (an Beispiele)
    • führen ein Modell aus
    • kennen wissen- und datenbasierten Modellierung
    • kennen die Quellen von Wissen, Data und Unsicherheiten einer Domäne
    • kennen was Merkmale sind (die Charkteristische Eigenschaften/Parameter, welche ein Domain beschreiben). Die Merkmale  wurden erhalten aus gemessene Sensor Daten oder von berechnete Daten (aus Modelen), auch bekannt als "soft sensor data", oder aus kommunizierte (mobile) Daten.
    • Wie entstehen Hypothesen aus Merkmalen? - durch Kombination der Merkmalen/Variablen, welche eine Hypothese darstellen. 
          --> siehe Folie 9 in der Vorlesungsfolien : Bayesian Networks for Decision Making in Automotive Applications - see

                                                                                              https://moodle.th-ab.de/course/view.php?id=5539#section-18

     PDF File: 

    Lecture ( 2023-05-04 - ,...) Bayesian Networks for Cognitive & Object-orriented Modelling (Decision Making) in Automotive ApplicationsDatei 


           --> siehe unten: Erklärung Video zur Entscheidungsfindung durch Hypothesen aus Merkmale

    • können Sensor-Unsicherheiten modellieren und analysieren. 
    • Warum ist es wichtig die (Sensor) Unsicherheiten auch bei Entscheidungsfindung zu berücksichtigen? bzw. Was ist der Einfluss von Unsicherheiten auf der Schlussfolgerung?
    • einordnen welche Merkmale werden in die Model-Hypothesen verwendet?  Die Merkmale werden erhalten von gemessene Sensor Daten  oder von berechnete Daten  (auch bekannt als "soft sensor data")
    • Welche Datentypen werden fürs Modellieren/Lernen verwendet: Variablen Typ (boolean, numbered, interval, labeled) 
    • Was ist "Data Labeling"? z.B. true/false, bzw. in Beispiel: Manöver Klass (Überquerung der Spurmarkierung: Links/Rechts/Gerade)
    • Klassifizierung von Hypothesen (z.B. in einer Anwendungen wie Erkennung von Fahrmanöver) unter Unsicherheiten
    • Evidenz als Input in Bayes Netz (Entscheidungsfindung) Model und Interpretation von Klassifikation (Entscheidung) Ergebnis.
    • kennen Werkzeuge für Design, Simulation und Testen von Entscheidungsfindungsmodelle
    • einordnen für welche Anwendungen Entscheidungsfindung Methoden (wie Bayes Netze) angebracht und gut geeignet sind
             (diskutiert in der erste Vorlesung UND in eBook, Chapter 1: Introduction; 1.5 When to Use Probabilistic Nets?) 

    • In English: Note on Hypothesis (Cause) und Features (Effects) File
      Students must
      Mark as done

      Note on Hypothesis (Cause) und Features (Effects):

      A conclusion/decision is based on the Cause – Effect Relation.

      The Hypothesis is the diagnosis (i.e., the possible cause of a problem).

      How do we find that there is a problem? – Based on the symptoms of this problem, the Problem has measurable (or observable) consequences/effects.

      Therefore, we need for the decision/conclusion some features (German: Merkmale), i.e., measurements, and/or observations, which appear as consequences/effects of the Problems (causes).

      Thus, we ask ourselves (e.g., for the Burglary-Modell):

      What did we observe/measure?

      • Was an Alarm issued/sounding?
      • Did Watson call due to the alarm?
      • Did the Radio report a new message on earthquake?

      Based on the observed/measure features we conclude/decide on the causes of a certain problem (issued alarm): Burglary or Earthquake. This means, that we meet the decision on what was the possible cause (diagnosis) of a problem, based on the evidence (observations/ measurements).

       

      The HUGIN software tool is using the developed BN model to evaluate the hypotheses based on the concrete observations (evidences) of the features.

      Therefore, for the Burglary-Modell: 

      The Hypotheses are the possible causes of the problems:  Burglary or Earthquake,

      and the features are the observed effects (i.e., consequences or symptoms) of the problem and its evidences: Alarm, Watson Call, Radio-Message.

      --------------------------------------------------------

      Note:
             * The Hypotheses are generated from the Features. In other words: from the Features are contructed the Hypothesen.

             * The hypotheses are evaluated, based on the Evidence (=measured Features) and this allows to make the decision.

             The Hypotheses are used in two different types of decision tasks:

      1. Diagnosis:

                 For Diagnostics: die decision finds what Hypothesis has the highest probability, based on the observations/measurements (evidence). An example of Diagnostics task is the angina.oobn Bayes Network.

            2. Prognosis:  

      Different Hypotheses are combined in order to build the prognosis um eine Prognosis, e.g. for the Recognition of maneuvers, based on observed features.

      In the example " Recognition of maneuvers": The decision (=Maneuver Recognition) has several states: DONTCARE, OBJCUTIN, OBJCUTOUT, LANEFOLLOW, OBJFOLLOW, EGOOCUTIN, EGOCUTOUT. 
      These states are computed based on the evidence (= observed features).
      For example, the Lateral Evidence is a Hypothesis, which is build from the features (lateral velocity and laterale distance (offset).
      The features are measured and set/entered as evidence in the Hypothesis.
      A measured value (e.g. speed) is influenced by the really measured Variable (velocity) und by the uncertainty/inacuracy of measurement/Noise. 

          The decision finds what combination of Hypotheses (with the corresponding features) - in what state - has the highest probability. The ranking of states according to their probabilities allows to set priority on different states of the decision. 

    • Examples of BN to practice for your presentation// Einige Beispiele (Bayes Netze) zum Üben File
      Students must
      Mark as done

      To download the BN examples (click on the icon).

      I have collected here some examples of BN to help your preparation for the exam. You can select one of these examples or you can also present your own example (or another known example).

      During the exam - you can explain the relevant exam questions based on your selected BN. 

      Sucessfull Preparations for the exam! 


      • //  (In German: Einige Bayes Netzwerke zu verschiedene Entscheidungsfindung Beispiele.

        Sie können ein Bayes Netz aus den eingefügten Beispiele wählen oder ein eigenen (oder ein bekannten)  Beispiel verwenden.

      • Bei der Prüfung können Sie die Prüfungsfragen anhand eines Beispiels erläutern.

      •  Viel Erfolg bei der Vorbereitung!


      We can discuss all possible question about the exam at our last lecture+excercise  on 12 Januar.

      If needed, we can also arrange an aditional meeting. We talk on January 12th. about suitable times.


  • Note for the international students: see the english version of the explanation videos.

    • In German: Bemerkung über Hypothese (Ursache/Cause) und Merkmale (Wirkung/Effect) File
      Students must
      Mark as done

      Bemerkung über Hypothese (Ursache/Cause) und Merkmale (Wirkung/Effect):

      Eine Entscheidungsfindung basiert auf der Ursache – Wirkung Beziehung (= Cause – Effect Relation)

      Die Hypothese ist die Diagnose (d.h. die mögliche Ursachen eines Problems).

      Wie entdecken wir das es ein Problem gibt? – Anhand der Symptome dieses Problems, d.h. das Problem hat messbare (bzw. beobachtbare) Auswirkungen.

      Deshalb braucht man für die Entscheidung die Merkmale, d.h. Messungen oder Beobachtungen, die als Folge des Problems (Ursache) erscheinen.

      Also, wir fragen uns (z.B. bei dem Burglary-Modell):

      Was hat man beobachtet?

      • Hat sich ein Alarm ausgelöst?
      • Hat Watson angerufen deswegen?
      • Hat das Radio neue Nachrichten zur Lage (=Erdbeben) berichtet?

      Anhand der Merkmale schließen wir zurück auf die Ursache eines Problems (Alarm Auslösung): Einbruch oder Erdbeben, d.h. wir treffen die Entscheidung über was war die mögliche Ursache (=Diagnose). Dabei werden die Hypothesen ausgewertet anhand der konkrete Beobachtungen (Evidenzen) der Merkmalen.

      Deshalb für das Burglary-Modell: 

      Die Hypothesen sind die Mögliche Ursachen des Problems:  Einbruch oder Erdbeben,

      und die Merkmale sind die beobachtbare Auswirkungen (oder Folgen oder Symptome) des Problems: Alarm, Watson Anruf, Radio-Nachricht

      ---



    • In German: Erklärung Video zur Entscheidungsfindung durch Hypothesen aus Merkmale Page
      Students must
      Mark as done

               Erklärung:
             * Die Hypothesen werden aus den Merkmalen generiert. Oder anders gesagt: aus den Merkmalen entstehen die Hypothesen.

             * Basierend auf den Evidenz (=gemessene Merkmalle) werden die Hypothesen ausgewertet und somit findet man die Entscheidung.

             Die Hypothesen werden in zwei unterschiedliche Typen von Entscheidungsfindung-Aufgaben verwendet:

      1. Diagnose:

                 Bei Diagnostik: die Entscheidung findet welche Hypothese die größte Wahrscheinlichkeit hat, basierend auf die Beobachtungen/Messungen (=Evidenzen). Ein Diagnostik Beispiel ist angina.oobn Bayes Netz.

            2. Prognose:  

      Verschiedene Hypothesen werden kombiniert um eine Prognose zu bilden wie z.B. bei der Erkennung von Manöver anhand von beobachteten Merkmale.
      In dem Beispiel mit Erkennung von Manöver: die Zustände der Entscheidung (=Manövererkennung) sind DONTCARE, OBJCUTIN, OBJCUTOUT, LANEFOLLOW, OBJFOLLOW, EGOOCUTIN, EGOCUTOUT. Diese werden anhand der beobachteten Merkmale (= Evidenzen) berechnet.
      Beispielsweise, die Laterale Evidenz ist eine Hypothese, welche aus den Merkmalen (laterale Geschwindigkeit und laterale Abstand (offset) gebildet wird.
      Die Merkmale werden gemessen und als Evidenzen in der Hypothese eingetragen.
      Ein gemessener Wert (z.B. Geschwindigkeit) wird beeinflusst von der tatsächlich gemessene Variable (Geschwindigkeit) und von der Fehler (=Unsicherheit wegen Messrauschen/Noise) bei der Messung. 

          Die Entscheidung findet welche Kombination von Hypothesen (mit den entsprechenden Merkmale) - für welche Zustand  - die größte Wahrscheinlichkeit hat und diese höchste Wahrscheinlichkeit ergibt die Entscheidung.