@article{13371, abstract = {Diamondoid nanoporous crystals represent a synthetically challenging class of materials that typically have been obtained from tetrahedral building blocks. In this issue of Chem, Stoddart and coworkers demonstrate that it is possible to generate diamondoid frameworks from a hexacationic building block lacking a tetrahedral symmetry. These results highlight the great potential of self-assembly for rapidly transforming small molecules into structurally complex functional materials.}, author = {Białek, Michał J. and Klajn, Rafal}, issn = {2451-9294}, journal = {Chem}, keywords = {Materials Chemistry, Biochemistry (medical), General Chemical Engineering, Environmental Chemistry, Biochemistry, General Chemistry}, number = {9}, pages = {2283--2285}, publisher = {Elsevier}, title = {{Diamond grows up}}, doi = {10.1016/j.chempr.2019.08.012}, volume = {5}, year = {2019}, } @article{13372, abstract = {The capacity to respond or adapt to environmental changes is an intrinsic property of living systems that comprise highly-connected subcomponents communicating through chemical networks. The development of responsive synthetic systems is a relatively new research area that covers different disciplines, among which nanochemistry brings conceptually new demonstrations. Especially attractive are ligand-protected gold nanoparticles, which have been extensively used over the last decade as building blocks in constructing superlattices or dynamic aggregates, under the effect of an applied stimulus. To reflect the importance of surface chemistry and nanoparticle core composition in the dynamic self-assembly of nanoparticles, we provide here an overview of various available stimuli, as tools for synthetic chemists to exploit. Along with this task, the review starts with the use of chemical stimuli such as solvent, pH, gases, metal ions or biomolecules. It then focuses on physical stimuli: temperature, magnetic and electric fields, as well as light. To reflect on the increasing complexity of current architectures, we discuss systems that are responsive to more than one stimulus, to finally encourage further research by proposing future challenges.}, author = {Grzelczak, Marek and Liz-Marzán, Luis M. and Klajn, Rafal}, issn = {1460-4744}, journal = {Chemical Society Reviews}, keywords = {General Chemistry}, number = {5}, pages = {1342--1361}, publisher = {Royal Society of Chemistry}, title = {{Stimuli-responsive self-assembly of nanoparticles}}, doi = {10.1039/c8cs00787j}, volume = {48}, year = {2019}, } @article{13373, abstract = {The reversible photoisomerization of azobenzene has been utilized to construct a plethora of systems in which optical, electronic, catalytic, and other properties can be controlled by light. However, owing to azobenzene’s hydrophobic nature, most of these examples have been realized only in organic solvents, and systems operating in water are relatively scarce. Here, we show that by coadsorbing the inherently hydrophobic azobenzenes with water-solubilizing ligands on the same nanoparticulate platforms, it is possible to render them essentially water-soluble. To this end, we developed a modified nanoparticle functionalization procedure allowing us to precisely fine-tune the amount of azobenzene on the functionalized nanoparticles. Molecular dynamics simulations helped us to identify two distinct supramolecular architectures (depending on the length of the background ligand) on these nanoparticles, which can explain their excellent aqueous solubilities. Azobenzenes adsorbed on these water-soluble nanoparticles exhibit highly reversible photoisomerization upon exposure to UV and visible light. Importantly, the mixed-monolayer approach allowed us to systematically investigate how the background ligand affects the switching properties of azobenzene. We found that the nature of the background ligand has a profound effect on the kinetics of azobenzene switching. For example, a hydroxy-terminated background ligand is capable of accelerating the back-isomerization reaction by more than 6000-fold. These results pave the way toward the development of novel light-responsive nanomaterials operating in aqueous media and, in the long run, in biological environments.}, author = {Chu, Zonglin and Han, Yanxiao and Bian, Tong and De, Soumen and Král, Petr and Klajn, Rafal}, issn = {1520-5126}, journal = {Journal of the American Chemical Society}, keywords = {Colloid and Surface Chemistry, Biochemistry, General Chemistry, Catalysis}, number = {5}, pages = {1949--1960}, publisher = {American Chemical Society}, title = {{Supramolecular control of azobenzene switching on nanoparticles}}, doi = {10.1021/jacs.8b09638}, volume = {141}, year = {2019}, } @article{13468, abstract = {Hydrogen-rich supernovae, known as Type II (SNe II), are the most common class of explosions observed following the collapse of the core of massive stars. We used analytical estimates and population synthesis simulations to assess the fraction of SNe II progenitors that are expected to have exchanged mass with a companion prior to explosion. We estimate that 1/3 to 1/2 of SN II progenitors have a history of mass exchange with a binary companion before exploding. The dominant binary channels leading to SN II progenitors involve the merger of binary stars. Mergers are expected to produce a diversity of SN II progenitor characteristics, depending on the evolutionary timing and properties of the merger. Alternatively, SN II progenitors from interacting binaries may have accreted mass from their companion, and subsequently been ejected from the binary system after their companion exploded. We show that the overall fraction of SN II progenitors that are predicted to have experienced binary interaction is robust against the main physical uncertainties in our models. However, the relative importance of different binary evolutionary channels is affected by changing physical assumptions. We further discuss ways in which binarity might contribute to the observed diversity of SNe II by considering potential observational signatures arising from each binary channel. For supernovae which have a substantial H-rich envelope at explosion (i.e., excluding Type IIb SNe), a surviving non-compact companion would typically indicate that the supernova progenitor star was in a wide, non-interacting binary. We argue that a significant fraction of even Type II-P SNe are expected to have gained mass from a companion prior to explosion.}, author = {Zapartas, Emmanouil and de Mink, Selma E. and Justham, Stephen and Smith, Nathan and de Koter, Alex and Renzo, Mathieu and Arcavi, Iair and Farmer, Rob and Götberg, Ylva Louise Linsdotter and Toonen, Silvia}, issn = {1432-0746}, journal = {Astronomy & Astrophysics}, keywords = {Space and Planetary Science, Astronomy and Astrophysics}, publisher = {EDP Sciences}, title = {{The diverse lives of progenitors of hydrogen-rich core-collapse supernovae: The role of binary interaction}}, doi = {10.1051/0004-6361/201935854}, volume = {631}, year = {2019}, } @article{13469, abstract = {Stars stripped of their envelopes from interaction with a binary companion emit a significant fraction of their radiation as ionizing photons. They are potentially important stellar sources of ionizing radiation, however, they are still often neglected in spectral synthesis simulations or simulations of stellar feedback. In anticipating the large datasets of galaxy spectra from the upcoming James Webb Space Telescope, we modeled the radiative contribution from stripped stars by using detailed evolutionary and spectral models. We estimated their impact on the integrated spectra and specifically on the emission rates of H I-, He I-, and He II-ionizing photons from stellar populations. We find that stripped stars have the largest impact on the ionizing spectrum of a population in which star formation halted several Myr ago. In such stellar populations, stripped stars dominate the emission of ionizing photons, mimicking a younger stellar population in which massive stars are still present. Our models also suggest that stripped stars have harder ionizing spectra than massive stars. The additional ionizing radiation, with which stripped stars contribute affects observable properties that are related to the emission of ionizing photons from stellar populations. In co-eval stellar populations, the ionizing radiation from stripped stars increases the ionization parameter and the production efficiency of hydrogen ionizing photons. They also cause high values for these parameters for about ten times longer than what is predicted for massive stars. The effect on properties related to non-ionizing wavelengths is less pronounced, such as on the ultraviolet continuum slope or stellar contribution to emission lines. However, the hard ionizing radiation from stripped stars likely introduces a characteristic ionization structure of the nebula, which leads to the emission of highly ionized elements such as O2+ and C3+. We, therefore, expect that the presence of stripped stars affects the location in the BPT diagram and the diagnostic ratio of O III to O II nebular emission lines. Our models are publicly available through CDS database and on the STARBURST99 website.}, author = {Götberg, Ylva Louise Linsdotter and de Mink, S. E. and Groh, J. H. and Leitherer, C. and Norman, C.}, issn = {1432-0746}, journal = {Astronomy & Astrophysics}, keywords = {Space and Planetary Science, Astronomy and Astrophysics}, publisher = {EDP Sciences}, title = {{The impact of stars stripped in binaries on the integrated spectra of stellar populations}}, doi = {10.1051/0004-6361/201834525}, volume = {629}, year = {2019}, } @article{13470, abstract = {Context. Massive Wolf–Rayet (WR) stars dominate the radiative and mechanical energy budget of galaxies and probe a critical phase in the evolution of massive stars prior to core collapse. It is not known whether core He-burning WR stars (classical WR; cWR) form predominantly through wind stripping (w-WR) or binary stripping (b-WR). Whereas spectroscopy of WR binaries has so-far largely been avoided because of its complexity, our study focuses on the 44 WR binaries and binary candidates of the Large Magellanic Cloud (LMC; metallicity Z ≈ 0.5 Z⊙), which were identified on the basis of radial velocity variations, composite spectra, or high X-ray luminosities. Aims. Relying on a diverse spectroscopic database, we aim to derive the physical and orbital parameters of our targets, confronting evolution models of evolved massive stars at subsolar metallicity and constraining the impact of binary interaction in forming these stars. Methods. Spectroscopy was performed using the Potsdam Wolf–Rayet (PoWR) code and cross-correlation techniques. Disentanglement was performed using the code Spectangular or the shift-and-add algorithm. Evolutionary status was interpreted using the Binary Population and Spectral Synthesis (BPASS) code, exploring binary interaction and chemically homogeneous evolution. Results. Among our sample, 28/44 objects show composite spectra and are analyzed as such. An additional five targets show periodically moving WR primaries but no detected companions (SB1); two (BAT99 99 and 112) are potential WR + compact-object candidates owing to their high X-ray luminosities. We cannot confirm the binary nature of the remaining 11 candidates. About two-thirds of the WN components in binaries are identified as cWR, and one-third as hydrogen-burning WR stars. We establish metallicity-dependent mass-loss recipes, which broadly agree with those recently derived for single WN stars, and in which so-called WN3/O3 stars are clear outliers. We estimate that 45  ±  30% of the cWR stars in our sample have interacted with a companion via mass transfer. However, only ≈12  ±  7% of the cWR stars in our sample naively appear to have formed purely owing to stripping via a companion (12% b-WR). Assuming that apparently single WR stars truly formed as single stars, this comprises ≈4% of the whole LMC WN population, which is about ten times less than expected. No obvious differences in the properties of single and binary WN stars, whose luminosities extend down to log L ≈ 5.2 [L⊙], are apparent. With the exception of a few systems (BAT99 19, 49, and 103), the equatorial rotational velocities of the OB-type companions are moderate (veq ≲ 250 km s−1) and challenge standard formalisms of angular-momentum accretion. For most objects, chemically homogeneous evolution can be rejected for the secondary, but not for the WR progenitor. Conclusions. No obvious dichotomy in the locations of apparently single and binary WN stars on the Hertzsprung-Russell diagram is apparent. According to commonly used stellar evolution models (BPASS, Geneva), most apparently single WN stars could not have formed as single stars, implying that they were stripped by an undetected companion. Otherwise, it must follow that pre-WR mass-loss/mixing (e.g., during the red supergiant phase) are strongly underestimated in standard stellar evolution models.}, author = {Shenar, T. and Sablowski, D. P. and Hainich, R. and Todt, H. and Moffat, A. F. J. and Oskinova, L. M. and Ramachandran, V. and Sana, H. and Sander, A. A. C. and Schnurr, O. and St-Louis, N. and Vanbeveren, D. and Götberg, Ylva Louise Linsdotter and Hamann, W.-R.}, issn = {1432-0746}, journal = {Astronomy & Astrophysics}, keywords = {Space and Planetary Science, Astronomy and Astrophysics}, publisher = {EDP Sciences}, title = {{The Wolf–Rayet binaries of the nitrogen sequence in the Large Magellanic Cloud}}, doi = {10.1051/0004-6361/201935684}, volume = {627}, year = {2019}, } @article{13471, abstract = {We perform an extensive numerical study of the evolution of massive binary systems to predict the peculiar velocities that stars obtain when their companion collapses and disrupts the system. Our aim is to (i) identify which predictions are robust against model uncertainties and assess their implications, (ii) investigate which physical processes leave a clear imprint and may therefore be constrained observationally, and (iii) provide a suite of publicly available model predictions to allow for the use of kinematic constraints from the Gaia mission. We find that 22+26−8% of all massive binary systems merge prior to the first core-collapse in the system. Of the remainder, 86+11−9% become unbound because of the core-collapse. Remarkably, this rarely produces runaway stars (observationally defined as stars with velocities above 30 km s−1). These are outnumbered by more than an order of magnitude by slower unbound companions, or “walkaway stars”. This is a robust outcome of our simulations and is due to the reversal of the mass ratio prior to the explosion and widening of the orbit, as we show analytically and numerically. For stars more massive than 15 M⊙, we estimate that 10+5−8% are walkaways and only 0.5+1.0−0.4% are runaways, nearly all of which have accreted mass from their companion. Our findings are consistent with earlier studies; however, the low runaway fraction we find is in tension with observed fractions of about 10%. Thus, astrometric data on presently single massive stars can potentially constrain the physics of massive binary evolution. Finally, we show that the high end of the mass distributions of runaway stars is very sensitive to the assumed black hole natal kicks, and we propose this as a potentially stringent test for the explosion mechanism. We also discuss companions remaining bound that can evolve into X-ray and gravitational wave sources.}, author = {Renzo, M. and Zapartas, E. and de Mink, S. E. and Götberg, Ylva Louise Linsdotter and Justham, S. and Farmer, R. J. and Izzard, R. G. and Toonen, S. and Sana, H.}, issn = {1432-0746}, journal = {Astronomy & Astrophysics}, keywords = {Space and Planetary Science, Astronomy and Astrophysics}, publisher = {EDP Sciences}, title = {{Massive runaway and walkaway stars}}, doi = {10.1051/0004-6361/201833297}, volume = {624}, year = {2019}, } @article{13472, abstract = {Massive stars in binaries can give rise to extreme phenomena such as X-ray binaries and gravitational wave sources after one or both stars end their lives as core-collapse supernovae. Stars in close orbit around a stellar or compact companion are expected to explode as “stripped-envelope supernovae”, showing no (Type Ib/c) or little (Type IIb) signs of hydrogen in the spectra, because hydrogen-rich progenitors are too large to fit. The physical processes responsible for the stripping process and the fate of the companion are still very poorly understood. Aiming to find new clues, we investigate Cas A, which is a very young (∼340 yr) and near (∼3.4 kpc) remnant of a core-collapse supernova. Cas A has been subject to several searches for possible companions, all unsuccessfully. We present new measurements of the proper motions and photometry of stars in the vicinity based on deep HST ACS/WFC and WFC3-IR data. We identify stellar sources that are close enough in projection but using their proper motions we show that none are compatible with being at the location of center at the time of explosion, in agreement with earlier findings. Our photometric measurements allow us to place much deeper (order-of-magnitude) upper limits on the brightness of possible undetected companions. We systematically compare them with model predictions for a wide variety of scenarios. We can confidently rule out the presence of any stellar companion of any reasonable mass and age (main sequence, pre main sequence or stripped) ruling out what many considered to be likely evolutionary scenarios for Type IIb supernova (SN IIb). More exotic scenarios that predict the presence of a compact companion (white dwarf, neutron star or black hole) are still possible as well as scenarios where the progenitor of Cas A was single at the moment of explosion (either because it was truly single, or resulted from a binary that was disrupted, or from a binary merger). The presence of a compact companion would imply that Cas A is of interest to study exotic outcomes of binary evolution. The single-at-death solution would still require fine-tuning of the process that removed most of the envelope through a mass-loss mechanism yet to be identified. We discuss how future constraints from Gaia and even deeper photometric studies may help to place further constraints.}, author = {Kerzendorf, Wolfgang E. and Do, Tuan and de Mink, Selma E. and Götberg, Ylva Louise Linsdotter and Milisavljevic, Dan and Zapartas, Emmanouil and Renzo, Mathieu and Justham, Stephen and Podsiadlowski, Philipp and Fesen, Robert A.}, issn = {1432-0746}, journal = {Astronomy & Astrophysics}, keywords = {Space and Planetary Science, Astronomy and Astrophysics}, publisher = {EDP Sciences}, title = {{No surviving non-compact stellar companion to Cassiopeia A}}, doi = {10.1051/0004-6361/201732206}, volume = {623}, year = {2019}, } @article{138, abstract = {Autoregulation is the direct modulation of gene expression by the product of the corresponding gene. Autoregulation of bacterial gene expression has been mostly studied at the transcriptional level, when a protein acts as the cognate transcriptional repressor. A recent study investigating dynamics of the bacterial toxin–antitoxin MazEF system has shown how autoregulation at both the transcriptional and post-transcriptional levels affects the heterogeneity of Escherichia coli populations. Toxin–antitoxin systems hold a crucial but still elusive part in bacterial response to stress. This perspective highlights how these modules can also serve as a great model system for investigating basic concepts in gene regulation. However, as the genomic background and environmental conditions substantially influence toxin activation, it is important to study (auto)regulation of toxin–antitoxin systems in well-defined setups as well as in conditions that resemble the environmental niche.}, author = {Nikolic, Nela}, journal = {Current Genetics}, number = {1}, pages = {133--138}, publisher = {Springer}, title = {{Autoregulation of bacterial gene expression: lessons from the MazEF toxin–antitoxin system}}, doi = {10.1007/s00294-018-0879-8}, volume = {65}, year = {2019}, } @article{14001, abstract = {Chiral molecules interact and react differently with other chiral objects, depending on their handedness. Therefore, it is essential to understand and ultimately control the evolution of molecular chirality during chemical reactions. Although highly sophisticated techniques for the controlled synthesis of chiral molecules have been developed, the observation of chirality on the natural femtosecond time scale of a chemical reaction has so far remained out of reach in the gas phase. Here, we demonstrate a general experimental technique, based on high-harmonic generation in tailored laser fields, and apply it to probe the time evolution of molecular chirality during the photodissociation of 2-iodobutane. These measurements show a change in sign and a pronounced increase in the magnitude of the chiral response over the first 100 fs, followed by its decay within less than 500 fs, revealing the photodissociation to achiral products. The observed time evolution is explained in terms of the variation of the electric and magnetic transition-dipole moments between the lowest electronic states of the cation as a function of the reaction coordinate. These results open the path to investigations of the chirality of molecular-reaction pathways, light-induced chirality in chemical processes, and the control of molecular chirality through tailored laser pulses.}, author = {Baykusheva, Denitsa Rangelova and Zindel, Daniel and Svoboda, Vít and Bommeli, Elias and Ochsner, Manuel and Tehlar, Andres and Wörner, Hans Jakob}, issn = {1091-6490}, journal = {Proceedings of the National Academy of Sciences}, keywords = {Multidisciplinary}, number = {48}, pages = {23923--23929}, publisher = {Proceedings of the National Academy of Sciences}, title = {{Real-time probing of chirality during a chemical reaction}}, doi = {10.1073/pnas.1907189116}, volume = {116}, year = {2019}, } @inproceedings{14002, abstract = {The advancement of attosecond chronoscopy has made it possible to reveal ultrashort time dynamics of photoionization [1]. Ionization delay measurements in atomic targets provide a wealth of information about the timing of the photoelectric effect [2], resonances, electron correlations and transport. The extension of this approach to molecules, however, presents great challenges. In addition to the difficulty of identifying correct ionization channels, it is hard to disentangle the role of the anisotropic molecular landscape from the delays inherent to the excitation process itself. Here, we present the measurements of ionization delays from ethyl iodide around the 4d giant dipole resonance of iodine. We employ attosecond streaking spectroscopy, which enables to disentangle the contribution to the delay from the functional ethyl group, being responsible for the characteristic chemical reactivity of the molecule. An attosecond extreme ultraviolet (XUV) pulse ionizes the molecule around the energy of the giant resonance and the released electron is exposed to the ponderomotive force of a synchronized near-infrared (NIR) field, which yields a streaking spectrogram (see figure). Comparative phase analysis of the spectrograms corresponding to iodine 4d and neon 2p emission permits extracting overall photoemission delays for ethyl iodide. The data is recorded for multiple photon energies around the iodine 4d resonance and compared to classical Wigner propagation [3] and quantum scattering [4] calculations. Here the outgoing electron, produced via inner shell ionization of the iodine atom in ethyl iodide, and thereby hardly influenced by the molecular potential during the birth process, acquires the necessary information about the influence of the functional ethyl group during its propagation. We find significant delay contributions that can distinguish between different functional groups, providing a sensitive probe of the local molecular environment [5]. This would stimulate to perform further angle resolved measurements in molecules to probe the potential landscape in three dimension.}, author = {Biswas, Shubhadeep and Liontos, I. and Kamal, A. M. and Kling, N. G. and Alharbi, A. F. and Alharbi, M. and Azzeer, A. M. and Worner, H. J. and Landsman, A. S. and Kling, M. F. and Forg, B. and Schotz, J. and Schweinberger, W. and Ortmann, L. and Zimmermann, T. and Pi, L.-W. and Baykusheva, Denitsa Rangelova and Masood, H. A.}, booktitle = {2019 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference}, isbn = {9781728104706}, location = {Munich, Germany}, publisher = {Institute of Electrical and Electronics Engineers}, title = {{Probing molecular influence on photoemission delays}}, doi = {10.1109/cleoe-eqec.2019.8871819}, year = {2019}, } @inproceedings{14184, abstract = {Learning disentangled representations is considered a cornerstone problem in representation learning. Recently, Locatello et al. (2019) demonstrated that unsupervised disentanglement learning without inductive biases is theoretically impossible and that existing inductive biases and unsupervised methods do not allow to consistently learn disentangled representations. However, in many practical settings, one might have access to a limited amount of supervision, for example through manual labeling of (some) factors of variation in a few training examples. In this paper, we investigate the impact of such supervision on state-of-the-art disentanglement methods and perform a large scale study, training over 52000 models under well-defined and reproducible experimental conditions. We observe that a small number of labeled examples (0.01--0.5\% of the data set), with potentially imprecise and incomplete labels, is sufficient to perform model selection on state-of-the-art unsupervised models. Further, we investigate the benefit of incorporating supervision into the training process. Overall, we empirically validate that with little and imprecise supervision it is possible to reliably learn disentangled representations.}, author = {Locatello, Francesco and Tschannen, Michael and Bauer, Stefan and Rätsch, Gunnar and Schölkopf, Bernhard and Bachem, Olivier}, booktitle = {8th International Conference on Learning Representations}, location = {Virtual}, title = {{Disentangling factors of variation using few labels}}, year = {2019}, } @inproceedings{14189, abstract = {We consider the problem of recovering a common latent source with independent components from multiple views. This applies to settings in which a variable is measured with multiple experimental modalities, and where the goal is to synthesize the disparate measurements into a single unified representation. We consider the case that the observed views are a nonlinear mixing of component-wise corruptions of the sources. When the views are considered separately, this reduces to nonlinear Independent Component Analysis (ICA) for which it is provably impossible to undo the mixing. We present novel identifiability proofs that this is possible when the multiple views are considered jointly, showing that the mixing can theoretically be undone using function approximators such as deep neural networks. In contrast to known identifiability results for nonlinear ICA, we prove that independent latent sources with arbitrary mixing can be recovered as long as multiple, sufficiently different noisy views are available.}, author = {Gresele, Luigi and Rubenstein, Paul K. and Mehrjou, Arash and Locatello, Francesco and Schölkopf, Bernhard}, booktitle = {Proceedings of the 35th Conference on Uncertainty in Artificial Intelligence}, location = {Tel Aviv, Israel}, pages = {217--227}, publisher = {ML Research Press}, title = {{The incomplete Rosetta Stone problem: Identifiability results for multi-view nonlinear ICA}}, volume = {115}, year = {2019}, } @inproceedings{14190, abstract = {Learning meaningful and compact representations with disentangled semantic aspects is considered to be of key importance in representation learning. Since real-world data is notoriously costly to collect, many recent state-of-the-art disentanglement models have heavily relied on synthetic toy data-sets. In this paper, we propose a novel data-set which consists of over one million images of physical 3D objects with seven factors of variation, such as object color, shape, size and position. In order to be able to control all the factors of variation precisely, we built an experimental platform where the objects are being moved by a robotic arm. In addition, we provide two more datasets which consist of simulations of the experimental setup. These datasets provide for the first time the possibility to systematically investigate how well different disentanglement methods perform on real data in comparison to simulation, and how simulated data can be leveraged to build better representations of the real world. We provide a first experimental study of these questions and our results indicate that learned models transfer poorly, but that model and hyperparameter selection is an effective means of transferring information to the real world.}, author = {Gondal, Muhammad Waleed and Wüthrich, Manuel and Miladinović, Đorđe and Locatello, Francesco and Breidt, Martin and Volchkov, Valentin and Akpo, Joel and Bachem, Olivier and Schölkopf, Bernhard and Bauer, Stefan}, booktitle = {Advances in Neural Information Processing Systems}, isbn = {9781713807933}, location = {Vancouver, Canada}, title = {{On the transfer of inductive bias from simulation to the real world: a new disentanglement dataset}}, volume = {32}, year = {2019}, } @inproceedings{14191, abstract = {A broad class of convex optimization problems can be formulated as a semidefinite program (SDP), minimization of a convex function over the positive-semidefinite cone subject to some affine constraints. The majority of classical SDP solvers are designed for the deterministic setting where problem data is readily available. In this setting, generalized conditional gradient methods (aka Frank-Wolfe-type methods) provide scalable solutions by leveraging the so-called linear minimization oracle instead of the projection onto the semidefinite cone. Most problems in machine learning and modern engineering applications, however, contain some degree of stochasticity. In this work, we propose the first conditional-gradient-type method for solving stochastic optimization problems under affine constraints. Our method guarantees O(k−1/3) convergence rate in expectation on the objective residual and O(k−5/12) on the feasibility gap.}, author = {Locatello, Francesco and Yurtsever, Alp and Fercoq, Olivier and Cevher, Volkan}, booktitle = {Advances in Neural Information Processing Systems}, isbn = {9781713807933}, location = {Vancouver, Canada}, pages = {14291–14301}, title = {{Stochastic Frank-Wolfe for composite convex minimization}}, volume = {32}, year = {2019}, } @inproceedings{14193, abstract = {A disentangled representation encodes information about the salient factors of variation in the data independently. Although it is often argued that this representational format is useful in learning to solve many real-world down-stream tasks, there is little empirical evidence that supports this claim. In this paper, we conduct a large-scale study that investigates whether disentangled representations are more suitable for abstract reasoning tasks. Using two new tasks similar to Raven's Progressive Matrices, we evaluate the usefulness of the representations learned by 360 state-of-the-art unsupervised disentanglement models. Based on these representations, we train 3600 abstract reasoning models and observe that disentangled representations do in fact lead to better down-stream performance. In particular, they enable quicker learning using fewer samples.}, author = {Steenkiste, Sjoerd van and Locatello, Francesco and Schmidhuber, Jürgen and Bachem, Olivier}, booktitle = {Advances in Neural Information Processing Systems}, isbn = {9781713807933}, location = {Vancouver, Canada}, title = {{Are disentangled representations helpful for abstract visual reasoning?}}, volume = {32}, year = {2019}, } @inproceedings{14197, abstract = {Recently there has been a significant interest in learning disentangled representations, as they promise increased interpretability, generalization to unseen scenarios and faster learning on downstream tasks. In this paper, we investigate the usefulness of different notions of disentanglement for improving the fairness of downstream prediction tasks based on representations. We consider the setting where the goal is to predict a target variable based on the learned representation of high-dimensional observations (such as images) that depend on both the target variable and an \emph{unobserved} sensitive variable. We show that in this setting both the optimal and empirical predictions can be unfair, even if the target variable and the sensitive variable are independent. Analyzing the representations of more than \num{12600} trained state-of-the-art disentangled models, we observe that several disentanglement scores are consistently correlated with increased fairness, suggesting that disentanglement may be a useful property to encourage fairness when sensitive variables are not observed.}, author = {Locatello, Francesco and Abbati, Gabriele and Rainforth, Tom and Bauer, Stefan and Schölkopf, Bernhard and Bachem, Olivier}, booktitle = {Advances in Neural Information Processing Systems}, isbn = {9781713807933}, location = {Vancouver, Canada}, pages = {14611–14624}, title = {{On the fairness of disentangled representations}}, volume = {32}, year = {2019}, } @inproceedings{14200, abstract = {The key idea behind the unsupervised learning of disentangled representations is that real-world data is generated by a few explanatory factors of variation which can be recovered by unsupervised learning algorithms. In this paper, we provide a sober look at recent progress in the field and challenge some common assumptions. We first theoretically show that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases on both the models and the data. Then, we train more than 12000 models covering most prominent methods and evaluation metrics in a reproducible large-scale experimental study on seven different data sets. We observe that while the different methods successfully enforce properties ``encouraged'' by the corresponding losses, well-disentangled models seemingly cannot be identified without supervision. Furthermore, increased disentanglement does not seem to lead to a decreased sample complexity of learning for downstream tasks. Our results suggest that future work on disentanglement learning should be explicit about the role of inductive biases and (implicit) supervision, investigate concrete benefits of enforcing disentanglement of the learned representations, and consider a reproducible experimental setup covering several data sets.}, author = {Locatello, Francesco and Bauer, Stefan and Lucic, Mario and Rätsch, Gunnar and Gelly, Sylvain and Schölkopf, Bernhard and Bachem, Olivier}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, location = {Long Beach, CA, United States}, pages = {4114--4124}, publisher = {ML Research Press}, title = {{Challenging common assumptions in the unsupervised learning of disentangled representations}}, volume = {97}, year = {2019}, } @article{14299, abstract = {DNA origami nano-objects are usually designed around generic single-stranded “scaffolds”. Many properties of the target object are determined by details of those generic scaffold sequences. Here, we enable designers to fully specify the target structure not only in terms of desired 3D shape but also in terms of the sequences used. To this end, we built design tools to construct scaffold sequences de novo based on strand diagrams, and we developed scalable production methods for creating design-specific scaffold strands with fully user-defined sequences. We used 17 custom scaffolds having different lengths and sequence properties to study the influence of sequence redundancy and sequence composition on multilayer DNA origami assembly and to realize efficient one-pot assembly of multiscaffold DNA origami objects. Furthermore, as examples for functionalized scaffolds, we created a scaffold that enables direct, covalent cross-linking of DNA origami via UV irradiation, and we built DNAzyme-containing scaffolds that allow postfolding DNA origami domain separation.}, author = {FAS, Engelhardt and Praetorius, Florian M and Wachauf, CH and Brüggenthies, G and Kohler, F and Kick, B and Kadletz, KL and Pham, PN and Behler, KL and Gerling, T and Dietz, H}, issn = {1936-086x}, journal = {ACS Nano}, number = {5}, pages = {5015--5027}, publisher = {ACS Publications}, title = {{Custom-size, functional, and durable DNA origami with design-specific scaffolds}}, doi = {10.1021/acsnano.9b01025}, volume = {13}, year = {2019}, } @article{17915, abstract = {The construction of self-assembled iron phthalocyanine (FePc) systems on gold electrodes modified by self-assembled monolayers (SAMs) is becoming an interesting strategy for obtaining electrocatalytic molecular building blocks for the oxygen reduction reaction (ORR). In this work, we have measured the conductance of pyridiniums axial ligands at the single molecule level using the scanning tunneling microscope-based break-junction method (STM-Break Junction) to study the role of the axial ligand on the activity of the self-assembled FePc systems on a gold electrode surface. The electron-pulling effect of pyridinium axial ligands is known to increase the electrocatalytic activity of FePc for the oxygen reduction reaction (ORR). We have used these systems as a platform for carrying out a comparative study for understanding the real influence of the proximal axial ligands. Further, these ligands act as molecular wires between the gold electrode surface and the FePc molecule. The pyridinium molecules were synthesized following a series of structural variations using a basic molecular backbone. From conductance measurements obtained for each pyridinium molecule, it was possible to establish that electron transport through each pyridinium does not influence the activity of FePc for ORR in alkaline media. In addition, the DFT calculations shows that the axial ligand in FePc modifies its catalytic activity by decreasing the binding energy of O2 to the Fe site.}, author = {Gutiérrez-Ceron, Cristian and Oñate, Rubén and Zagal, José H. and Pizarro, Ana and Silva, J. Francisco and Castro-Castillo, Carmen and Rezende, Marcos Caroli and Flores, Marcos and Cortés-Arriagada, Diego and Toro-Labbé, Alejandro and Campos, Luis M. and Venkataraman, Latha and Ponce, Ingrid}, issn = {0013-4686}, journal = {Electrochimica Acta}, publisher = {Elsevier}, title = {{Molecular conductance versus inductive effects of axial ligands on the electrocatalytic activity of self-assembled iron phthalocyanines: The oxygen reduction reaction}}, doi = {10.1016/j.electacta.2019.134996}, volume = {327}, year = {2019}, }