Algorithmc Foundations Of Data Science an der RWTH Aachen | Karteikarten & Zusammenfassungen

Lernmaterialien für Algorithmc Foundations of Data Science an der RWTH Aachen

Greife auf kostenlose Karteikarten, Zusammenfassungen, Übungsaufgaben und Altklausuren für deinen Algorithmc Foundations of Data Science Kurs an der RWTH Aachen zu.

TESTE DEIN WISSEN

Forms of Learning: Semi-Supervised learning  

Lösung anzeigen
TESTE DEIN WISSEN
  • set-up like supervised learning, but only few and possible faulty examples 
  • try to make the best out of the examples and data 
Lösung ausblenden
TESTE DEIN WISSEN

Forms of Learning: Unsupervised Learning

Lösung anzeigen
TESTE DEIN WISSEN
  • try to detect patterns in data 
  • no explicit feedback 
  • most important task is clustering
Lösung ausblenden
TESTE DEIN WISSEN

Forms of Learning: Supervised Learning

Lösung anzeigen
TESTE DEIN WISSEN
  • try to learn a function from examples 
  • we differentiate between: 
    • Classification: if the function is finite valued, we try to predict values for future input
    • Regression: if the fucntion is numerical, we try to predict expected values for future input.
Lösung ausblenden
TESTE DEIN WISSEN

Realisable Problems 

Lösung anzeigen
TESTE DEIN WISSEN

A learning Problem is realisable if the target function is in the hypothesis space. 

(we usually don't know this) 

Lösung ausblenden
TESTE DEIN WISSEN

Convergence of k-means algorithm 

Lösung anzeigen
TESTE DEIN WISSEN

The k-means algorithm always halts in a finite number of steps. 

Lösung ausblenden
TESTE DEIN WISSEN

Partitioning

Lösung anzeigen
TESTE DEIN WISSEN

Map Tasks: 

  • ususally more map tasks than nodes to improve load balancing
  • it is common to use on DFS chunk per map task 


Reduce Tasks

  • ususally fewer reduce tasks than map tasks
  • output of map phase often skewed, that is, value lists for different keys differ significantly in length 
  • randomly assigning keys to reduce taks usually reduces impact of skew. 
Lösung ausblenden
TESTE DEIN WISSEN

Underlying Assumption of the Nearest Neighbour Algorithm

Lösung anzeigen
TESTE DEIN WISSEN

Data items that are close together have similar function values, or belong to the same class for a classification task. 

Lösung ausblenden
TESTE DEIN WISSEN

Data Structures for Nearest Neighbour Classification 

Lösung anzeigen
TESTE DEIN WISSEN
  • Finding the nearest neighbours of a query point in s set of n data points requires O(n). For large data sets, this may be prohibitive (verboten).
  • We can reduce the query time by preprossing the data points into a suitable structure - k-d-trees (a version of binary search trees for k-dimensional data)
Lösung ausblenden
TESTE DEIN WISSEN

Forms of Learning: Reinforcement Learning 

Lösung anzeigen
TESTE DEIN WISSEN
  • try to detect actions that maximize reward
  • feedback is given as reward (or punishment) 
  • trial-and-error process
Lösung ausblenden
TESTE DEIN WISSEN

Occam's Razor 

Lösung anzeigen
TESTE DEIN WISSEN

Choose the simplest hypothesis consistent with the data. 

Lösung ausblenden
TESTE DEIN WISSEN

What are Realisable Learning Problems? 

Lösung anzeigen
TESTE DEIN WISSEN

A learning problem is realisable if the target function is in the hypothesis space 


(Usually we don't know this) 

Picking a good hypothesis space is part of the machine learning exercise 

Lösung ausblenden
TESTE DEIN WISSEN

Agnostic Learning 

Lösung anzeigen
TESTE DEIN WISSEN

A learning approach that does not assume the learning problem at hand is realisable. 

Lösung ausblenden
  • 177629 Karteikarten
  • 4280 Studierende
  • 318 Lernmaterialien

Beispielhafte Karteikarten für deinen Algorithmc Foundations of Data Science Kurs an der RWTH Aachen - von Kommilitonen auf StudySmarter erstellt!

Q:

Forms of Learning: Semi-Supervised learning  

A:
  • set-up like supervised learning, but only few and possible faulty examples 
  • try to make the best out of the examples and data 
Q:

Forms of Learning: Unsupervised Learning

A:
  • try to detect patterns in data 
  • no explicit feedback 
  • most important task is clustering
Q:

Forms of Learning: Supervised Learning

A:
  • try to learn a function from examples 
  • we differentiate between: 
    • Classification: if the function is finite valued, we try to predict values for future input
    • Regression: if the fucntion is numerical, we try to predict expected values for future input.
Q:

Realisable Problems 

A:

A learning Problem is realisable if the target function is in the hypothesis space. 

(we usually don't know this) 

Q:

Convergence of k-means algorithm 

A:

The k-means algorithm always halts in a finite number of steps. 

Mehr Karteikarten anzeigen
Q:

Partitioning

A:

Map Tasks: 

  • ususally more map tasks than nodes to improve load balancing
  • it is common to use on DFS chunk per map task 


Reduce Tasks

  • ususally fewer reduce tasks than map tasks
  • output of map phase often skewed, that is, value lists for different keys differ significantly in length 
  • randomly assigning keys to reduce taks usually reduces impact of skew. 
Q:

Underlying Assumption of the Nearest Neighbour Algorithm

A:

Data items that are close together have similar function values, or belong to the same class for a classification task. 

Q:

Data Structures for Nearest Neighbour Classification 

A:
  • Finding the nearest neighbours of a query point in s set of n data points requires O(n). For large data sets, this may be prohibitive (verboten).
  • We can reduce the query time by preprossing the data points into a suitable structure - k-d-trees (a version of binary search trees for k-dimensional data)
Q:

Forms of Learning: Reinforcement Learning 

A:
  • try to detect actions that maximize reward
  • feedback is given as reward (or punishment) 
  • trial-and-error process
Q:

Occam's Razor 

A:

Choose the simplest hypothesis consistent with the data. 

Q:

What are Realisable Learning Problems? 

A:

A learning problem is realisable if the target function is in the hypothesis space 


(Usually we don't know this) 

Picking a good hypothesis space is part of the machine learning exercise 

Q:

Agnostic Learning 

A:

A learning approach that does not assume the learning problem at hand is realisable. 

Algorithmc Foundations of Data Science

Erstelle und finde Lernmaterialien auf StudySmarter.

Greife kostenlos auf tausende geteilte Karteikarten, Zusammenfassungen, Altklausuren und mehr zu.

Jetzt loslegen

Das sind die beliebtesten Algorithmc Foundations of Data Science Kurse im gesamten StudySmarter Universum

Data Science

Universität Potsdam

Zum Kurs
Data Science

IU Internationale Hochschule

Zum Kurs

Die all-in-one Lernapp für Studierende

Greife auf Millionen geteilter Lernmaterialien der StudySmarter Community zu
Kostenlos anmelden Algorithmc Foundations of Data Science
Erstelle Karteikarten und Zusammenfassungen mit den StudySmarter Tools
Kostenlos loslegen Algorithmc Foundations of Data Science