Institute Talks

#### Metrics Matter, Examples from Binary and Multilabel Classification

IS Colloquium
• 21 August 2017 • 11:15 12:15
• Sanmi Koyejo
• Empirical Inference meeting room (MPI-IS building, 4th floor)

Performance metrics are a key component of machine learning systems, and are ideally constructed to reflect real world tradeoffs. In contrast, much of the literature simply focuses on algorithms for maximizing accuracy. With the increasing integration of machine learning into real systems, it is clear that accuracy is an insufficient measure of performance for many problems of interest. Unfortunately, unlike accuracy, many real world performance metrics are non-decomposable i.e. cannot be computed as a sum of losses for each instance. Thus, known algorithms and associated analysis are not trivially extended, and direct approaches require expensive combinatorial optimization. I will outline recent results characterizing population optimal classifiers for large families of binary and multilabel classification metrics, including such nonlinear metrics as F-measure and Jaccard measure. Perhaps surprisingly, the prediction which maximizes the utility for a range of such metrics takes a simple form. This results in simple and scalable procedures for optimizing complex metrics in practice. I will also outline how the same analysis gives optimal procedures for selecting point estimates from complex posterior distributions for structured objects such as graphs. Joint work with Nagarajan Natarajan, Bowei Yan, Kai Zhong, Pradeep Ravikumar and Inderjit Dhillon.

Organizers: Mijung Park

#### Physical Blendshapes - Controllable Physics for Human Faces

Talk
• 23 August 2017 • 11:00 12:00
• Yeara Kozlov
• Aquarium

Creating convincing human facial animation is challenging. Face animation is often hand-crafted by artists separately from body motion. Alternatively, if the face animation is derived from motion capture, it is typically performed while the actor is relatively still. Recombining the isolated face animation with body motion is non-trivial and often results in uncanny results if the body dynamics are not properly reflected on the face (e.g. cheeks wiggling when running). In this talk, I will discuss the challenges of human soft tissue simulation and control. I will then present our method for adding physical effects to facial blendshape animation. Unlike previous methods that try to add physics to face rigs, our method can combine facial animation and rigid body motion consistently while preserving the original animation as closely as possible. Our novel simulation framework uses the original animation as per-frame rest-poses without adding spurious forces. We also propose the concept of blendmaterials to give artists an intuitive means to control the changing material properties due to muscle activation.

Organizers: Timo Bolkart

#### Dominik Bach - TBA

IS Colloquium
• 02 October 2017 • 11:15 12:15
• Dominik Bach

#### A Collection of Recent Work: From 6D Pose estimation via MRF-Diversity to zebrafish detection

Talk
• 21 November 2016 • 11:30 12:15
• Carsten Rother
• MRZ seminar room

In this talk I will present the portfolio of work we conduct in our lab. Herby, I will present three recent body of work in more detail. This is firstly our work on learning 6D Object Pose estimation and Camera localizing from RGB or RGBD images. I will show that by utilizing the concepts of uncertainty and learning to score hypothesis, we can improve the state of the art. Secondly, I will present a new approach for inferring multiple diverse labeling in a graphical model. Besides guarantees of an exact solution, our method is also faster than existing techniques. Finally, I will present a recent work in which we show that popular Auto-context Decision Forests can be mapped to Deep ConvNets for Semantic Segmentation. We use this to detect the spine of a zebrafish, in case when little training data is available.

Organizers: Aseem Behl

#### Future of graphical models: more modeling power, parallelization, scalable solvers.

Talk
• 21 November 2016 • 12:15 12:45
• Dr. Bogdan Savchynskyy
• MRZ seminar room

We propose a new computational framework for combinatorial problems arising in machine learning and computer vision. This framework is a special case of Lagrangean (dual) decomposition, but allows for efficient dual ascent (message passing) optimization. In a sense, one can understand both the framework and the optimization technique as a generalization of those for standard undirected graphical models (conditional random fields). We will make an overview of our recent results and plans for the nearest future.

Organizers: Aseem Behl

#### Future of graphical models: more modeling power, parallelization, scalable solvers

Talk
• 21 November 2016 • 12:15 13:00
• Bogdan Savchynskyy
• Mrz Seminar Room (room no. 0.A.03)

We propose a new computational framework for combinatorial problems arising in machine learning and computer vision. This framework is a special case of Lagrangean (dual) decomposition, but allows for efficient dual ascent (message passing) optimization. In a sense, one can understand both the framework and the optimization technique as a generalization of those for standard undirected graphical models (conditional random fields). We will make an overview of our recent results and plans for the nearest future.

Organizers: Aseem Behl

#### Monte Carlo with determinantal point processes

Talk
• 02 November 2016 • 15:00 16:00
• Rémi Bardenet
• AGBS Seminar room (Spemannstr. 38)

In this talk, we show that using repulsive random variables, it is possible to build Monte Carlo methods that converge faster than vanilla Monte Carlo. More precisely, we build estimators of integrals, the variance of which decreases as $N^{-1-1/d}$, where $N$ is the number of integrand evaluations, and $d$ is the ambient dimension. To do so, we propose stochastic numerical quadratures involving determinantal point processes (DPPs) associated to multivariate orthogonal polynomials. The proposed method can be seen as a stochastic version of Gauss' quadrature, where samples from a determinantal point process replace zeros of orthogonal polynomials. Furthermore, integration with DPPs is close in spirit to randomized quasi-Monte Carlo methods, leveraging repulsive point processes to ensure low discrepancy samples. The talk is based on the following preprint https://arxiv.org/abs/1605.00361

Organizers: Alexandra Gessner

#### Factorized Latent Representations for Improved Automated Diagnostics

Talk
• 27 October 2016 • 15:00 16:00
• Hedvig Kjellström
• MRZ Seminar Room

In this talk I will first outline my different research projects. I will then focus on one project with applications in Health, and introduce the Inter-Battery Topic Model (IBTM). Our approach extends traditional topic models by learning a factorized latent variable representation. The structured representation leads to a model that marries benefits traditionally associated with a discriminative approach, such as feature selection, with those of a generative model, such as principled regularization and ability to handle missing data. The factorization is provided by representing data in terms of aligned pairs of observations as different views. This provides means for selecting a representation that separately models topics that exist in both views from the topics that are unique to a single view. This structured consolidation allows for efficient and robust inference and provides a compact and efficient representation.

#### Optical Robot Skin and Whole Body Vision

Talk
• 19 October 2016 • 14:00 15:00
• Chris Atkeson and Akihiko Yamaguchi
• Max Planck House, Lecture Hall

Chris Atkeson will talk about the motivation for optical robot skin and whole-body vision. Akihiko Yamaguchi will talk about a first application, FingerVision.

Organizers: Ludovic Righetti

#### Numerics in Computational Stellar Astrophysics

Talk
• 29 September 2016 • 14:00 15:00
• Jean-Claude Passy
• AGBS Seminar room (Spemmanstr. 38)

The importance of computer science in astrophysical research has increased tremendously over the past 15 years. Indeed, as observational facilities and missions are constantly pushing their precision limit, theorists need to provide observers with more and more realistic numerical models. These models need to be verified, validated, and their uncertainties must be assessed. In this talk, I will present the results of two independent numerical studies aiming at solving some fundamental problems in stellar astrophysics. First, I will explain how we have used different 3D hydrodynamics codes to simulate stellar mergers. In particular I will focus on the verification and validation steps, and describe a new algorithm to compute self-gravity that I have developed and implemented in a grid-based code. Then, I will introduce the concept of a ‘stellar evolution' code which models the full evolution of a star, from its birth until its death. I will present a code comparison of several such codes widely used by the astrophysical community, and assess their systematic uncertainties. These modeling uncertainties must be taken into account by observers if they wish to derive observed parameters more reliably.

Organizers: Raffi Enficiaud

#### Considering uncertainty in robot decision-making: control and modelling aspects

Talk
• 27 September 2016 • 11:00 12:00
• Jose R. Medina
• AMD Seminar Room (Paul-Ehrlich-Str. 15, 1rst floor)

Control under uncertainty is an omnipresent problem in robotics that typically arises when robots must cope with unknown environments/tasks. Robot control typically ignores uncertainty by considering only the expected outcomes of the robot’s internal model. Interestingly, neuroscientist have shown that humans adapt their decisions depending on the level of uncertainty which is not reflected in the expected values, but in higher order statistics. In this talk I will first present an approach to systematically address this problem in the context of stochastic optimal control. I will then give an example of how the robot’s internal model structure defines the level uncertainty and its distribution. Finally, experiments in a physical human-robot interaction setting will illustrate the capabilities of this approach.

Organizers: Ludovic Righetti

#### Multi-contact Stability: Support Areas and Volumes for Humanoid Locomotion under Frictional Contacts

Talk
• 19 September 2016 • 2:00 p.m. 3:00 p.m.
• Stéphane Caron
• AMD Seminar Room (Paul-Ehrlich-Str. 15, 1rst floor)

Humanoid locomotion on horizontal floors was solved by closing the feedback loop on the Zero-tiling Moment Point (ZMP), a measurable dynamic point that needs to stay inside the foot contact area to prevent the robot from falling (contact stability criterion). However, this criterion does not apply to general multi-contact settings, the "new frontier" in humanoid locomotion. In this talk, we will see how the ideas of ZMP and support area can be generalized and applied to multi-contact locomotion. First, we will show how support areas can be calculated in any virtual plane, allowing one to apply classical schemes even when contacts are not coplanar. Yet, these schemes constraint the center-of-mass (COM) to planar motions. We overcome this limitation by extending the calculation of the contact-stability criterion from a support area to a support cone of 3D COM accelerations. We use this new criterion to implement a multi-contact walking pattern generator based on predictive control of COM accelerations, which we will demonstrate in real-time simulations during the presentation.

Organizers: Ludovic Righetti

#### Graph decomposition for multi-person tracking, pose estimation and motion segmentation

Talk
• 25 August 2016 • 11:00 12:00
• Siyu Tang
• MRZ Seminar Room (Spemannstr 41)

Understanding people in images and videos is a problem studied intensively in computer vision. While continuous progress has been made, occlusions, cluttered background, complex poses and large variety of appearance remain challenging, especially for crowded scenes. In this talk, I will explore the algorithms and tools that enable computer to interpret people's position, motion and articulated poses in the real-world challenging images and videos.More specifically, I will discuss an optimization problem whose feasible solutions define a decomposition of a given graph. I will highlight the applications of this problem in computer vision, which range from multi-person tracking [1,2,3] to motion segmentation [4]. I will also cover an extended optimization problem whose feasible solutions define a decomposition of a given graph and a labeling of its nodes with the application on multi-person pose estimation [5]. Reference: [1] Subgraph Decomposition for Multi-Object Tracking; S. Tang, B. Andres, M. Andriluka and B. Schiele; CVPR 2015 [2] Multi-Person Tracking by Multicut and Deep Matching; S. Tang, B. Andres, M. Andriluka and B. Schiele; arXiv 2016 [3] Multi-Person Tracking by Lifted Multicut and Person Re-identification; S. Tang, B. Andres, M. Andriluka and B. Schiele [4] A Multi-cut Formulation for Joint Segmentation and Tracking of Multiple Objects; M. Keuper, S. Tang, Z. Yu, B. Andres, T. Brox and B. Schiele; arXiv 2016 [5] DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation.: L. Pishchulin, E. Insafutdinov, S. Tang, B. Andres, M. Andriluka, P. Gehler and B. Schiele; CVPR16

Organizers: Naureen Mahmood