Knowledge Discovery at Karlsruher Institut für Technologie

Flashcards and summaries for Knowledge Discovery at the Karlsruher Institut für Technologie

Arrow Arrow

It’s completely free

studysmarter schule studium
d

4.5 /5

studysmarter schule studium
d

4.8 /5

studysmarter schule studium
d

4.5 /5

studysmarter schule studium
d

4.8 /5

Study with flashcards and summaries for the course Knowledge Discovery at the Karlsruher Institut für Technologie

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

What is supervised learning?

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:


Which methods for splitting training and test data for evaluation do exist? (5x)


Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Splitting methods - Hold-out 

  • Explanation
  • (Dis-)Advantages
This was only a preview of our StudySmarter flashcards.
Flascard Icon Flascard Icon

Millions of flashcards created by students

Flascard Icon Flascard Icon

Create your own flashcards as quick as possible

Flascard Icon Flascard Icon

Learning-Assistant with spaced repetition algorithm

Sign up for free!

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Splitting methods - Stratified sampling

  • Explanation
  • (Dis-)Advantages

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

How can Precision and Recall be improved?

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Dealing with overfitting - Strategies 

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

How can overfitting in decision trees be avoided? (2 Approaches)

This was only a preview of our StudySmarter flashcards.
Flascard Icon Flascard Icon

Millions of flashcards created by students

Flascard Icon Flascard Icon

Create your own flashcards as quick as possible

Flascard Icon Flascard Icon

Learning-Assistant with spaced repetition algorithm

Sign up for free!

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

What describes Linear Separability?

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

How can we optimize Multi-Layer ANNs?

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Optimization Algorithms - AdaGrad

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Optimization Algorithms - RMSProp

This was only a preview of our StudySmarter flashcards.
Flascard Icon Flascard Icon

Millions of flashcards created by students

Flascard Icon Flascard Icon

Create your own flashcards as quick as possible

Flascard Icon Flascard Icon

Learning-Assistant with spaced repetition algorithm

Sign up for free!

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Optimization Algorithms - Adam

Your peers in the course Knowledge Discovery at the Karlsruher Institut für Technologie create and share summaries, flashcards, study plans and other learning materials with the intelligent StudySmarter learning app.

Get started now!

Flashcard Flashcard

Exemplary flashcards for Knowledge Discovery at the Karlsruher Institut für Technologie on StudySmarter:

Knowledge Discovery

What is supervised learning?

Goal is to learn a function that specifies the output for given input data

Knowledge Discovery


Which methods for splitting training and test data for evaluation do exist? (5x)


  1. Hold-out
  2. k-fold cross-validation
  3. Leave-one-out
  4. Bootstrapping
  5. Stratified sampling

Knowledge Discovery

Splitting methods - Hold-out 

  • Explanation
  • (Dis-)Advantages

Explanation

  • Split available data into two subsets: Training and Test Set
  • Train on training data, evaluate on test data

Advantages

  • Evaluation on data which is not used for training
  • easy to implement

Disadvantages

  • Not all the data is used for training
  • Evaluation results strongly depend on choice of training and test set

Knowledge Discovery

Splitting methods - Stratified sampling

  • Explanation
  • (Dis-)Advantages

Explanation

  • To "preserve" (= erhalten) characteristics of data

Advantage

  • Ability to preserve from unlabeled data

Disadvantages

  • Very complex
  • Requires high skills

Knowledge Discovery

How can Precision and Recall be improved?

  • Usually Precision leads to low Recall and vice versa
    • high Precision --> less FP & a lot TP --> less FN --> low Recall


Improve Precision: Return only relevant documents / test data

--> predicts cancer only if confident


Improve Recall: Return all documents / test data!

--> we do not miss too many cases of cancer

Knowledge Discovery

Dealing with overfitting - Strategies 

Strategy 1

  • use more training data

Strategy 2

  • controll complexity
  • For complex models, there is a greater chance that they are fitted accidentally by errors in data
  • Try to choose less complex model class from the beginning or control the complexity of the resulting

    model during learning

  • Use one simpel and one complex model


Conclusion 

  • Eine Verallgemeinerung ist möglich, wenn ein Modell auf einer ausreichend großen Trainingsdatenmenge gut funktioniert und es nicht

    zu komplex


Knowledge Discovery

How can overfitting in decision trees be avoided? (2 Approaches)

Preprunning

  • Stoppen Sie die Baumkonstruktion frühzeitig - teilen Sie einen Knoten nicht auf, wenn dies dazu führen würde, dass das Gutheitsmaß unter den Schwellenwert fällt
  • Difficult to choose an appropriate threshold 
  • Approach
    • Stop if number of instances is less than some user-specified threshold
    • Stop if expanding the current node does not improve information content

Postprunning

  • Entfernen Sie Äste von einem "ausgewachsenen" Baum - erhalten Sie eine Folge von progressiv beschnittenen Bäumen
  • Approach
    • Grow decision tree to its entirety
    • Trim the nodes of the decision tree in a bottom-up fashion

Knowledge Discovery

What describes Linear Separability?

Linear Separability – Classification problems for which a model can be expressed in the form of linear threshold functions are called linearly separable.


-->  Lineare Trennbarkeit -Klassifikationsprobleme, für die ein Modell in Form von linearen Schwellenwertfunktionen ausgedrückt werden kann, werden als linear trennbar.

Knowledge Discovery

How can we optimize Multi-Layer ANNs?

  • Dropout
  • Optimization Algorithms
    • Gradient Descent with Momentum
    • AdaGrad
    • RMSProp
    • Adam

Knowledge Discovery

Optimization Algorithms - AdaGrad

  • AdaGrad adapts the learning rate to the parameters
    • die Durchführung kleinerer Updates für Parameter, die mit häufig auftretenden Merkmalen verbunden sind, und
    • größere Updates für Parameter, die mit seltenen Features verbunden sind
  • Good for sparse (spärliche) data

Knowledge Discovery

Optimization Algorithms - RMSProp

  • AdaGrad decays the learning rate very aggressively
    • As a result the parameters will start receiving very small updates because of the decayed learning rate.
  • RMSProp automatically adjusts the learning rate

Knowledge Discovery

Optimization Algorithms - Adam

  • Adam is the combination of Momentum and RMSProp
    • Acceleration SGD in relevant direction, and
    • automatically update of learning rate 

Sign up for free to see all flashcards and summaries for Knowledge Discovery at the Karlsruher Institut für Technologie

Singup Image Singup Image
Wave

Other courses from your degree program

For your degree program Knowledge Discovery at the Karlsruher Institut für Technologie there are already many courses on StudySmarter, waiting for you to join them. Get access to flashcards, summaries, and much more.

Back to Karlsruher Institut für Technologie overview page

Discovering Chemistry

General Knowledge

Data Mining and Knowledge Discovery

Information & Knowledge Management

Knowledge Management

What is StudySmarter?

What is StudySmarter?

StudySmarter is an intelligent learning tool for students. With StudySmarter you can easily and efficiently create flashcards, summaries, mind maps, study plans and more. Create your own flashcards e.g. for Knowledge Discovery at the Karlsruher Institut für Technologie or access thousands of learning materials created by your fellow students. Whether at your own university or at other universities. Hundreds of thousands of students use StudySmarter to efficiently prepare for their exams. Available on the Web, Android & iOS. It’s completely free.

Awards

Best EdTech Startup in Europe

Awards
Awards

EUROPEAN YOUTH AWARD IN SMART LEARNING

Awards
Awards

BEST EDTECH STARTUP IN GERMANY

Awards
Awards

Best EdTech Startup in Europe

Awards
Awards

EUROPEAN YOUTH AWARD IN SMART LEARNING

Awards
Awards

BEST EDTECH STARTUP IN GERMANY

Awards
X

StudySmarter - The study app for students

StudySmarter

4.5 Stars 1100 Rating
Start now!
X

Good grades at university? No problem with StudySmarter!

89% of StudySmarter users achieve better grades at university.

50 Mio Flashcards & Summaries
Create your own content with Smart Tools
Individual Learning-Plan

Learn with over 1 million users on StudySmarter.

Already registered? Just go to Login