Seminar on Internet Technologies (Summer 2016): Difference between revisions

Jump to navigation Jump to search
Line 103: Line 103:
*[http://tools.ietf.org/pdf/draft-pentikousis-icn-scenarios-04.pdf  ICN Base line scenarios]
*[http://tools.ietf.org/pdf/draft-pentikousis-icn-scenarios-04.pdf  ICN Base line scenarios]
|-
|-
| '''Learning from Imbalanced Data'''   
| '''Learning from Imbalanced Data (assigned to Georgios Kaklamanos)'''   
When building and training classifiers for classification problems, one commonly encountered problem is that of imbalanced data. For instance, in the case of a binary classifier, this means that one class is hugely overrepresented in the data available. Training classifiers for this kind of datasets has been a problem for some time. In this work, your task is to i) precisely introduce the imbalanced data problem, ii) discuss the state of the art of approaches for mitigating this problem (both from the perspective of learning algorithms and data manipulation techniques) and iii) find out what issues still remain open until today. Note that this topic requires a background in data science, and in particular in classification algorithms. Also, this topic requires a comparatively high reading effort.
When building and training classifiers for classification problems, one commonly encountered problem is that of imbalanced data. For instance, in the case of a binary classifier, this means that one class is hugely overrepresented in the data available. Training classifiers for this kind of datasets has been a problem for some time. In this work, your task is to i) precisely introduce the imbalanced data problem, ii) discuss the state of the art of approaches for mitigating this problem (both from the perspective of learning algorithms and data manipulation techniques) and iii) find out what issues still remain open until today. Note that this topic requires a background in data science, and in particular in classification algorithms. Also, this topic requires a comparatively high reading effort.
| [https://www.net.informatik.uni-goettingen.de/people/David_Koll David Koll ]
| [https://www.net.informatik.uni-goettingen.de/people/David_Koll David Koll ]
| [http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=5128907&tag=1]
| [http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=5128907&tag=1]
|-
|-
| '''Deep Learning and its (possible) flaws'''   
| '''Deep Learning and its (possible) flaws (assigned to Christoph Rauterberg)'''   
One recent trend in machine learning is 'deep learning', where neural networks are employed for solving a wide range of problems. One prominent example of such problems is image classification. While neural networks are in fact delivering sometimes great results, they may also have some weak spots. In this work, your task is to i) make yourself familiar with neural networks, ii) discuss the state-of-the-art in image classification, and iii) to investigate some possible flaws in neural networks. Note that for this topic a background in data science, and in particular in classification algorithms, is strongly recommended. Also, this topic requires a comparatively high reading effort.
One recent trend in machine learning is 'deep learning', where neural networks are employed for solving a wide range of problems. One prominent example of such problems is image classification. While neural networks are in fact delivering sometimes great results, they may also have some weak spots. In this work, your task is to i) make yourself familiar with neural networks, ii) discuss the state-of-the-art in image classification, and iii) to investigate some possible flaws in neural networks. Note that for this topic a background in data science, and in particular in classification algorithms, is strongly recommended. Also, this topic requires a comparatively high reading effort.
| [https://www.net.informatik.uni-goettingen.de/people/David_Koll David Koll ]
| [https://www.net.informatik.uni-goettingen.de/people/David_Koll David Koll ]

Navigation menu