Filters








25,657 Hits in 3.6 sec

Vienne, Paris, Hambourg... Werner Hofmann et l'histoire de l'art

Werner Hofmann, Éric Darragon, Pierre Georgel, Dario Gamboni, Thomas Gaehtgens
2007 Perspective  
Alors quel est le vrai Werner Hofmann ?  ...  et s'est achevée avec Goya, l'ère des révolutions Perspective, 3 | 2007 Thomas Gaehtgens. Je souhaiterais également commencer par un souvenir.  ... 
doi:10.4000/perspective.3596 fatcat:xmgc22jgf5cjlewwzauf3u6nmu

Generator Reversal [article]

Yannic Kilcher, Aurélien Lucchi, Thomas Hofmann
2017 arXiv   pre-print
We consider the problem of training generative models with deep neural networks as generators, i.e. to map latent codes to data points. Whereas the dominant paradigm combines simple priors over codes with complex deterministic models, we propose instead to use more flexible code distributions. These distributions are estimated non-parametrically by reversing the generator map during training. The benefits include: more powerful generative models, better modeling of latent structure and explicit control of the degree of generalization.
arXiv:1707.09241v1 fatcat:rf55bbjs3ngbzle2sc5cynhmwu

:{unav)

Thomas Hofmann
2012 Machine Learning  
Aspects versus clusters It is worth comparing the aspect model with statistical clustering models (cf. also Hofmann, Puzicha, & Jordan, 1999) .  ...  The combination of deterministic annealing with the EM algorithm has been investigated before in Ueda and Nakano (1998), Hofmann, Puzicha, and Jordan (1999) .  ... 
doi:10.1023/a:1007617005950 fatcat:gt5wfjjr3jefzmxhavjtognziy

How to Query Language Models? [article]

Leonard Adolphs, Shehzaad Dhuliawala, Thomas Hofmann
2021 arXiv   pre-print
., 2021; Adolphs and Hofmann, 2019), or-as in the experiment at hand-tidying up a room (Murugesan et al., 2020).  ... 
arXiv:2108.01928v1 fatcat:kifflzbmfbg3hlwpys56dyvc54

Autoregressive Text Generation Beyond Feedback Loops [article]

Florian Schmidt, Stephan Mandt, Thomas Hofmann
2019 arXiv   pre-print
Further, we restrict our model to unary potentials to obtain a non-autoregressive state space model similar to that of Schmidt and Hofmann (2018) , denoted SSM.  ...  Non-autoregressive sequence models have recently regained attention for unconditional (Schmidt and Hofmann, 2018; M. Ziegler and M. Rush, 2019) and conditional (Lee et al., 2018) generation.  ... 
arXiv:1908.11658v1 fatcat:myrn5rldmfhhdg2t7tptat5beq

Mixing of Stochastic Accelerated Gradient Descent [article]

Peiyuan Zhang, Hadi Daneshmand, Thomas Hofmann
2019 arXiv   pre-print
[3] Hadi Daneshmand, Jonas Kohler, Aurelien Luc- chi, and Thomas Hofmann. Escaping sad- dles with stochastic gradients. arXiv preprint arXiv:1803.05999, 2018.  ... 
arXiv:1910.14616v1 fatcat:yzd574a3hzdgljzyqbmayusc3a

Accelerated Dual Learning by Homotopic Initialization [article]

Hadi Daneshmand, Hamed Hassani, Thomas Hofmann
2017 arXiv   pre-print
Gradient descent and coordinate descent are well understood in terms of their asymptotic behavior, but less so in a transient regime often used for approximations in machine learning. We investigate how proper initialization can have a profound effect on finding near-optimal solutions quickly. We show that a certain property of a data set, namely the boundedness of the correlations between eigenfeatures and the response variable, can lead to faster initial progress than expected by commonplace
more » ... nalysis. Convex optimization problems can tacitly benefit from that, but this automatism does not apply to their dual formulation. We analyze this phenomenon and devise provably good initialization strategies for dual optimization as well as heuristics for the non-convex case, relevant for deep learning. We find our predictions and methods to be experimentally well-supported.
arXiv:1706.03958v1 fatcat:igca3q77xvdutnzxdtuix2bfd4

Hyperbolic Neural Networks [article]

Octavian-Eugen Ganea, Gary Bécigneul, Thomas Hofmann
2018 arXiv   pre-print
Hyperbolic spaces have recently gained momentum in the context of machine learning due to their high capacity and tree-likeliness properties. However, the representational power of hyperbolic geometry is not yet on par with Euclidean geometry, mostly because of the absence of corresponding hyperbolic neural network layers. This makes it hard to use hyperbolic embeddings in downstream tasks. Here, we bridge this gap in a principled manner by combining the formalism of M\"obius gyrovector spaces
more » ... ith the Riemannian geometry of the Poincar\'e model of hyperbolic spaces. As a result, we derive hyperbolic versions of important deep learning tools: multinomial logistic regression, feed-forward and recurrent neural networks such as gated recurrent units. This allows to embed sequential data and perform classification in the hyperbolic space. Empirically, we show that, even if hyperbolic optimization tools are limited, hyperbolic sentence embeddings either outperform or are on par with their Euclidean variants on textual entailment and noisy-prefix recognition tasks.
arXiv:1805.09112v2 fatcat:mwdkqqjeyna4fefx7xqtojjjna

Perioperative Infektionsprophylaxe in Unfallchirurgie und Orthopädie

Johann Pichl, Thomas Hofmann, Thomas Auhuber, Reinhard Hoffmann
2005 OP-Journal  
doi:10.1055/s-2007-977775 fatcat:xpzhrloyjjasnd36k44c4u6u44

Software Testing, AI and Robotics (STAIR) Learning Lab [article]

Simon Haller-Seeber, Thomas Gatterer, Patrick Hofmann, Christopher Kelter, Thomas Auer, Michael Felderer
2022 arXiv   pre-print
In this paper we presented the Software Testing, AI and Robotics (STAIR) Learning Lab. STAIR is an initiative started at the University of Innsbruck to bring robotics, Artificial Intelligence (AI) and software testing into schools. In the lab physical and virtual learning units are developed in parallel and in sync with each other. Its core learning approach is based the develop of both a physical and simulated robotics environment. In both environments AI scenarios (like traffic sign
more » ... n) are deployed and tested. We present and focus on our newly designed MiniBot that are both built on hardware which was designed for educational and research purposes as well as the simulation environment. Additionally, we describe first learning design concepts and a showcase scenario (i.e., AI-based traffic sign recognition) with different exercises which can easily be extended.
arXiv:2204.03028v1 fatcat:bft3fkvkmjb7hclivcohmh7ybm

DFS- ConceptDesk - Experimenteller Workspace für Fluglotsen [article]

Thomas Hofmann, Jörg Bergner
2016 Mensch & Computer  
Um die Arbeitsbelastung von Fluglotsen zu reduzieren wurde das DFS-ConceptDesk entwickelt. Dieses integrierte Hard- und Softwaresystem inkludiert zahlreiche technische Anzeige- und Interaktionssysteme in einem UX-Konzept. Das System besteht aus zwei Großflächendisplays, von denen das horizontale Display als Multitouch ausgerüstet ist. Das gesamte System wird über TouchInput und Tangibles bedient, die Kommunikation mit den Piloten kann darüber hinaus aus Sicherheitsgründen über Mikrophon
more » ... . Das experimentelle System integriert bisher separate HMI in einer homogenen Visualisierung und Interaktionsmethodik.
doi:10.18420/muc2016-up-0046 dblp:conf/mc/HofmannB16 fatcat:25d4pxtj2vfypnacd6tlzd6xai

Multiresistente Keime

Thomas Mückley, Michael Diefenbeck, Günther Hofmann
2005 OP-Journal  
Multiresistente Keime Thomas Mü ckley, Michael Diefenbeck, Gunther Olaf Hofmann Zusammenfassung Die aktuelle Prävalenz Methicillin-resistenter Staphylococcus -aureus-Stä mme (MRSA) von 20,7 % in Deutschland  ... 
doi:10.1055/s-2007-977780 fatcat:6rsyny4hnraz5kyto3gkd7qgpm

Probabilistic Latent Semantic Indexing

Thomas Hofmann
2017 SIGIR Forum  
doi:10.1145/3130348.3130370 fatcat:whi6wltfhrfj7ev35vwevu4spu

Transformative Wirtschaftswissenschaften?

Ulrich Petschow, Thomas Korbun, David Hofmann
2017 Ökologisches Wirtschaften - Fachzeitschrift  
Einführung in das Schwerpunktthema
doi:10.14512/oew320214 fatcat:yw7yuiasnzflfhjdewdqazzn5q

Controlling a d-level atom in a cavity [article]

Thomas Hofmann, Michael Keyl
2017 arXiv   pre-print
In this paper we study controllability of a d-level atom interacting with the electromagnetic field in a cavity. The system is modelled by an ordered graph Γ. The vertices of Γ describe the energy levels and the edges allowed transitions. To each edge of Γ we associate a harmonic oscillator representing one mode of the electromagnetic field. The dynamics of the system (drift) is given by a natural generalization of the Jaynes-Cummings Hamiltonian. If we add in addition sufficient control over
more » ... e atom, the overall system (atom and em-field) becomes strongly controllable, i.e. each unitary on the system Hilbert space can be approximated with arbitrary precision in the strong topology by control unitaries. A key role in the proof is played by a topological *-algebra which is (roughly speaking) a representation of the path algebra of Γ. It contains crucial structural information about the control problem, and is therefore an important tool for the implementation of control tasks like preparing a particular state from the ground state. This is demonstrated by a detailed discussion of different versions of three-level systems.
arXiv:1712.07613v1 fatcat:76zqigjpvvcqbhm77iwkwvtcvm
« Previous Showing results 1 — 15 out of 25,657 results