Forschungsinstitutionen

http://domino.mpi-inf.mpg.de/internet/events.nsf/0/AE7AB71943B3C069C1258704003690E0

Join Zoom Meeting
https://zoom.us/j/99715655535?pwd=ZnVlTXdrcUgwREo2S1NVcG5wcmY4QT09

Meeting ID: 997 1565 5535
Passcode: jls-21

---------------------------------------------------------------

Abstract:

In this presentation, I will talk about some of the
recent work we did on new methods for reconstructing
computer graphics models of real world scenes from
sparse or even monocular video data. These methods
are based of bringing together neural network-based
and explicit model-based approaches.

I will also talk about new neural rendering approaches
that combine explicit model-based and neural network
based concepts for image formation in new ways. They
enable new means to synthesize highly realistic
imagery and videos of real work scenes under user
control.

Presenter:

Christian Theobalt: https://people.mpi-inf.mpg.de/~theobalt/

--------------------------------------------------------------

Please find all speakers for this semester here:
https://nextcloud.mpi-klsb.mpg.de/index.php/s/LTNKiGWZ8e688kp

Abstract:
Natural language processing (NLP) plays an increasingly important role in everyday life, and many people are familiar with products such as Google Translate, Alexa or Siri. However, NLP systems currently only exist for a small fraction of the world's approximately 7000 languages. This is undesirable for many reasons: For instance, only speakers of high-resource languages are able to benefit from the abundance of information available on the internet, which reinforces already existing inequalities. It also limits the ability of NLP to support language documentation and revitalization efforts.
In my talk, I will present NLP research focused explicitly on low-resource languages. I will first talk about how we can leverage NLP systems to speed up language documentation efforts. I will then discuss how existing pretrained multilingual models can be adapted to truly low-resource languages.

Zoom-Meeting beitreten
https://zoom.us/j/96125745299?pwd=VXNkWVZBV0ZsdmNmSXFjRUpxZHZyQT09

Meeting-ID: 961 2574 5299
Kenncode: 636087
Schnelleinwahl mobil
+13017158592,,96125745299#,,,,*636087# Vereinigte Staaten (Washington DC)
+13126266799,,96125745299#,,,,*636087# Vereinigte Staaten (Chicago)

Einwahl nach aktuellem Standort
+1 301 715 8592 Vereinigte Staaten (Washington DC)
+1 312 626 6799 Vereinigte Staaten (Chicago)
+1 346 248 7799 Vereinigte Staaten (Houston)
+1 646 558 8656 Vereinigte Staaten (New York)
+1 669 900 9128 Vereinigte Staaten (San Jose)
+1 253 215 8782 Vereinigte Staaten (Tacoma)
Meeting-ID: 961 2574 5299
Kenncode: 636087
Ortseinwahl suchen: https://zoom.us/u/aiVAxyXnz

Join Zoom Meeting
https://zoom.us/j/99457028566?pwd=bGllck5vS05iYnhtMGVCUzFGS1JqUT09

Meeting ID: 994 5702 8566
Passcode: 780865

Abstract:
When using distributed machine learning (ML) systems to train models on
a cluster of worker machines, users must configure a large number of
parameters: hyper-parameters (e.g. the batch size and the learning rate)
affect model convergence; system parameters (e.g. the number of workers
and their communication topology) impact training performance. Some of
these parameters, such as the number of workers, may also change in
elastic machine learning scenarios. In current systems, adapting such
parameters during training is ill-supported.

In this talk, I will describe our recent work on KungFu, a distributed
deep learning library for TensorFlow and PyTorch that is designed to
enable adaptive and elastic training. KungFu allows users to express
high-level Adaptation Policies (APs) that describe how to change hyper-
and system parameters during training. APs take real-time monitored
metrics (e.g. signal-to-noise ratios) as input and trigger control
actions (e.g. cluster rescaling or synchronisation strategy updates).
For execution, APs are translated into monitoring and control operators
that are embedded in the dataflow graph. APs exploit an efficient
asynchronous collective communication layer, which ensures concurrency
and consistency of monitoring and adaptation operations.

Bio:
Peter Pietzuch is a Professor of Distributed Systems at Imperial College
London, where he leads the Large-scale Data & Systems (LSDS) group
(http://lsds.doc.ic.ac.uk). His research work focuses on the design and
engineering of scalable, reliable and secure large-scale software
systems, with a particular interest in performance, data management and
security issues. He has published papers in premier scientific venues,
including OSDI/SOSP, SIGMOD, VLDB, ASPLOS, USENIX ATC, EuroSys, SoCC,
ICDCS and Middleware. Currently he is a Visiting Researcher with
Microsoft Research and serves as the Director of Research in the
Department, the Chair of the ACM SIGOPS European Chapter, and an
Associate Editor for IEEE TKDE and TCC. Before joining Imperial College
London, he was a post-doctoral Fellow at Harvard University. He holds
PhD and MA degrees from the University of Cambridge.

https://www.dfki.de/web/news/detail/News/gemeinsam-frauen-in-der-ki-staerken-dfki-und-women-in-ai-robotics-starten-kooperation-mit-virtuellem/

Location: https://zoom.us/j/94577321297?pwd=N3l5K1ZtZ3E1aytnWlBkL1FUazNXZz09
Meeting ID: 945 7732 1297
Passcode: 205903

Hosts: Ignacio Cirac (MPQ) and Kurt Mehlhorn (MPI-INF).

Abstract: While there has been extensive previous work on efficient quantum algorithms for linear differential equations, analogous progress for nonlinear differential equations has been severely limited due to the linearity of quantum mechanics. Despite this obstacle, we develop a quantum algorithm for initial value problems described by dissipative quadratic n-dimensional ordinary differential equations. Assuming R<1, where R is a parameter characterizing the ratio of the nonlinearity to the linear dissipation, this algorithm has complexity T^2 poly(log T, log n, log(1/ϵ))/ϵ, where T is the evolution time and ϵ is the allowed error in the output quantum state. This is an exponential improvement over the best previous quantum algorithms, whose complexity is exponential in T. We achieve this improvement using the method of Carleman linearization, for which we give a novel convergence theorem. This method maps a system of nonlinear differential equations to an infinite-dimensional system of linear differential equations, which we discretize, truncate, and solve using the forward Euler method and the quantum linear system algorithm. We also provide a lower bound on the worst-case complexity of quantum algorithms for general quadratic differential equations, showing that the problem is intractable for R≥sqrt(2). Finally, we discuss potential applications of this approach to problems arising in biology as well as in fluid and plasma dynamics.

Based on joint work with Jin-Peng Liu, Herman Kolden, Hari Krovi, Nuno Loureiro, and Konstantina Trivisa.

http://domino.mpi-inf.mpg.de/internet/events.nsf/0/0FA99D7C57281E71C12587180026568B

Photorealistic and semantically controllable digital models of human faces are important for a wide range of applications such as movies, virtual reality, and casual photography. Traditional approaches require expensive setups which capture the person from multiple cameras under different illumination conditions. Recent approaches have also explored digitizing faces under less constrained settings, even from a single image of the person. These approaches rely on priors, commonly known as 3D morphable models (3DMMs), which are learned from datasets of 3D scans. This thesis pushes the state of the art in high-quality 3D reconstruction of faces from monocular images. A model-based face autoencoder architecture is introduced which integrates convolutional neural networks, 3DMMs, and differentiable rendering for self-supervised training on large image datasets. This architecture is extended to enable the refinement of a pretrained 3DMM just from a dataset of monocular images, allowing for higher-quality reconstructions. In addition, this thesis demonstrates the learning of the identity components of a 3DMM directly from videos without using any 3D data. Since videos are more readily available, this model can generalize better compared to the models learned from limited 3D scans.

https://www.graduateschool-computerscience.de/doctoral-students/314-2/

Computer vision systems deployed in the real-world will encounter inputs far from its training distribution. For example, a self-driving car might see a blue stop-sign it has not seen before. To ensure safe operation when faced with such out-of-distribution corner cases, it is vital to quantify the model robustness to such data before deployment. In this dissertation, we build generative models to create synthetic data variations at scale and leverage them to test the robustness of target computer vision systems to these variations. First, we build generative models which can controllably manipulate image and text data. This includes models to a) change visual context by removing objects, b) edit the appearance of an object in images, and c) change the writing style of text. Next, using these generative models we create model-agnostic and model-targeted input variations, to study the robustness of computer vision systems. While in the model-agnostic case, the input image is manipulated based on design priors, in the model-targeted case the manipulation is directly guided by the target model performance. With these methods, we measure and improve the robustness of various computer vision systems — specifically image classification, segmentation, object detection and visual question answering systems — to semantic input variations. Additionally, in the text domain, we deploy these generative models to improve diversity of image captioning systems and perform writing style manipulation to obfuscate private attributes of the user.

https://www.graduateschool-computerscience.de/doctoral-students/314-2/