Training a Neural Network in a Low-Resource Setting on Automatically Annotated Noisy Data
Jul 22, 2018
Michael A. Hedderich, Dietrich Klakow
Jul 22, 2018
Michael A. Hedderich, Dietrich Klakow




* In Proceedings of the Workshop on Deep Learning Approaches for Low-Resource NLP 2018
Click to Read Paper




* IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), New Orleans, LA, 2017, pp. 5710-5714
* Published at IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2017. arXiv admin note: text overlap with arXiv:1703.08068
Click to Read Paper
A Batch Noise Contrastive Estimation Approach for Training Large Vocabulary Language Models
Aug 22, 2017
Youssef Oualil, Dietrich Klakow
Aug 22, 2017
Youssef Oualil, Dietrich Klakow



* Accepted for publication at INTERSPEECH'17
Click to Read Paper
NEXUS Network: Connecting the Preceding and the Following in Dialogue Generation
Oct 07, 2018
Hui Su, Xiaoyu Shen, Wenjie Li, Dietrich Klakow
Oct 07, 2018
Hui Su, Xiaoyu Shen, Wenjie Li, Dietrich Klakow




* Accepted by EMNLP2018
Click to Read Paper
Long-Short Range Context Neural Networks for Language Modeling
Aug 22, 2017
Youssef Oualil, Mittul Singh, Clayton Greenberg, Dietrich Klakow
Aug 22, 2017
Youssef Oualil, Mittul Singh, Clayton Greenberg, Dietrich Klakow




* Published at EMNLP'16
Click to Read Paper
Sequential Recurrent Neural Networks for Language Modeling
Mar 23, 2017
Youssef Oualil, Clayton Greenberg, Mittul Singh, Dietrich Klakow
Mar 23, 2017
Youssef Oualil, Clayton Greenberg, Mittul Singh, Dietrich Klakow




* published (INTERSPEECH 2016), 5 pages, 3 figures, 4 tables
Click to Read Paper
Adversarial Initialization -- when your network performs the way I want
Feb 08, 2019
Kathrin Grosse, Thomas A. Trost, Marius Mosbach, Michael Backes, Dietrich Klakow
The increase in computational power and available data has fueled a wide deployment of deep learning in production environments. Despite their successes, deep architectures are still poorly understood and costly to train. We demonstrate in this paper how a simple recipe enables a market player to harm or delay the development of a competing product. Such a threat model is novel and has not been considered so far. We derive the corresponding attacks and show their efficacy both formally and empirically. These attacks only require access to the initial, untrained weights of a network. No knowledge of the problem domain and the data used by the victim is needed. On the initial weights, a mere permutation is sufficient to limit the achieved accuracy to for example 50% on the MNIST dataset or double the needed training time. While we can show straightforward ways to mitigate the attacks, the respective steps are not part of the standard procedure taken by developers so far.
Feb 08, 2019
Kathrin Grosse, Thomas A. Trost, Marius Mosbach, Michael Backes, Dietrich Klakow
* 16 pages, 20 figures
Click to Read Paper
Logit Pairing Methods Can Fool Gradient-Based Attacks
Oct 29, 2018
Marius Mosbach, Maksym Andriushchenko, Thomas Trost, Matthias Hein, Dietrich Klakow
Oct 29, 2018
Marius Mosbach, Maksym Andriushchenko, Thomas Trost, Matthias Hein, Dietrich Klakow




Click to Read Paper
Effective Slot Filling Based on Shallow Distant Supervision Methods
Jan 06, 2014
Benjamin Roth, Tassilo Barth, Michael Wiegand, Mittul Singh, Dietrich Klakow
Jan 06, 2014
Benjamin Roth, Tassilo Barth, Michael Wiegand, Mittul Singh, Dietrich Klakow




* to be published in: Proceedings of the Sixth Text Analysis Conference (TAC 2013)
Click to Read Paper