@article{6566, abstract = {Methodologies that involve the use of nanoparticles as “artificial atoms” to rationally build materials in a bottom-up fashion are particularly well-suited to control the matter at the nanoscale. Colloidal synthetic routes allow for an exquisite control over such “artificial atoms” in terms of size, shape, and crystal phase as well as core and surface compositions. We present here a bottom-up approach to produce Pb–Ag–K–S–Te nanocomposites, which is a highly promising system for thermoelectric energy conversion. First, we developed a high-yield and scalable colloidal synthesis route to uniform lead sulfide (PbS) nanorods, whose tips are made of silver sulfide (Ag2S). We then took advantage of the large surface-to-volume ratio to introduce a p-type dopant (K) by replacing native organic ligands with K2Te. Upon thermal consolidation, K2Te-surface modified PbS–Ag2S nanorods yield p-type doped nanocomposites with PbTe and PbS as major phases and Ag2S and Ag2Te as embedded nanoinclusions. Thermoelectric characterization of such consolidated nanosolids showed a high thermoelectric figure-of-merit of 1 at 620 K.}, author = {Ibáñez, Maria and Genç, Aziz and Hasler, Roger and Liu, Yu and Dobrozhan, Oleksandr and Nazarenko, Olga and Mata, María de la and Arbiol, Jordi and Cabot, Andreu and Kovalenko, Maksym V.}, issn = {1936-086X}, journal = {ACS Nano}, keywords = {colloidal nanoparticles, asymmetric nanoparticles, inorganic ligands, heterostructures, catalyst assisted growth, nanocomposites, thermoelectrics}, number = {6}, pages = {6572--6580}, publisher = {American Chemical Society}, title = {{Tuning transport properties in thermoelectric nanocomposites through inorganic ligands and heterostructured building blocks}}, doi = {10.1021/acsnano.9b00346}, volume = {13}, year = {2019}, } @inproceedings{6569, abstract = {Knowledge distillation, i.e. one classifier being trained on the outputs of another classifier, is an empirically very successful technique for knowledge transfer between classifiers. It has even been observed that classifiers learn much faster and more reliably if trained with the outputs of another classifier as soft labels, instead of from ground truth data. So far, however, there is no satisfactory theoretical explanation of this phenomenon. In this work, we provide the first insights into the working mechanisms of distillation by studying the special case of linear and deep linear classifiers. Specifically, we prove a generalization bound that establishes fast convergence of the expected risk of a distillation-trained linear classifier. From the bound and its proof we extract three keyfactors that determine the success of distillation: data geometry – geometric properties of the datadistribution, in particular class separation, has an immediate influence on the convergence speed of the risk; optimization bias– gradient descentoptimization finds a very favorable minimum of the distillation objective; and strong monotonicity– the expected risk of the student classifier always decreases when the size of the training set grows.}, author = {Bui Thi Mai, Phuong and Lampert, Christoph}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, location = {Long Beach, CA, United States}, pages = {5142--5151}, publisher = {ML Research Press}, title = {{Towards understanding knowledge distillation}}, volume = {97}, year = {2019}, } @article{6575, abstract = {Motivated by recent experimental observations of coherent many-body revivals in a constrained Rydbergatom chain, we construct a weak quasilocal deformation of the Rydberg-blockaded Hamiltonian, whichmakes the revivals virtually perfect. Our analysis suggests the existence of an underlying nonintegrableHamiltonian which supports an emergent SU(2)-spin dynamics within a small subspace of the many-bodyHilbert space. We show that such perfect dynamics necessitates the existence of atypical, nonergodicenergy eigenstates—quantum many-body scars. Furthermore, using these insights, we construct a toymodel that hosts exact quantum many-body scars, providing an intuitive explanation of their origin. Ourresults offer specific routes to enhancing coherent many-body revivals and provide a step towardestablishing the stability of quantum many-body scars in the thermodynamic limit.}, author = {Choi, Soonwon and Turner, Christopher J. and Pichler, Hannes and Ho, Wen Wei and Michailidis, Alexios and Papić, Zlatko and Serbyn, Maksym and Lukin, Mikhail D. and Abanin, Dmitry A.}, issn = {10797114}, journal = {Physical Review Letters}, number = {22}, publisher = {American Physical Society}, title = {{Emergent SU(2) dynamics and perfect quantum many-body scars}}, doi = {10.1103/PhysRevLett.122.220603}, volume = {122}, year = {2019}, } @article{6586, abstract = {The bottom-up assembly of colloidal nanocrystals is a versatile methodology to produce composite nanomaterials with precisely tuned electronic properties. Beyond the synthetic control over crystal domain size, shape, crystal phase, and composition, solution-processed nanocrystals allow exquisite surface engineering. This provides additional means to modulate the nanomaterial characteristics and particularly its electronic transport properties. For instance, inorganic surface ligands can be used to tune the type and concentration of majority carriers or to modify the electronic band structure. Herein, we report the thermoelectric properties of SnTe nanocomposites obtained from the consolidation of surface-engineered SnTe nanocrystals into macroscopic pellets. A CdSe-based ligand is selected to (i) converge the light and heavy bands through partial Cd alloying and (ii) generate CdSe nanoinclusions as a secondary phase within the SnTe matrix, thereby reducing the thermal conductivity. These SnTe-CdSe nanocomposites possess thermoelectric figures of merit of up to 1.3 at 850 K, which is, to the best of our knowledge, the highest thermoelectric figure of merit reported for solution-processed SnTe.}, author = {Ibáñez, Maria and Hasler, Roger and Genç, Aziz and Liu, Yu and Kuster, Beatrice and Schuster, Maximilian and Dobrozhan, Oleksandr and Cadavid, Doris and Arbiol, Jordi and Cabot, Andreu and Kovalenko, Maksym V.}, issn = {1520-5126}, journal = {Journal of the American Chemical Society}, number = {20}, pages = {8025--8029}, publisher = {American Chemical Society}, title = {{Ligand-mediated band engineering in bottom-up assembled SnTe nanocomposites for thermoelectric energy conversion}}, doi = {10.1021/jacs.9b01394}, volume = {141}, year = {2019}, } @inproceedings{6590, abstract = {Modern machine learning methods often require more data for training than a single expert can provide. Therefore, it has become a standard procedure to collect data from external sources, e.g. via crowdsourcing. Unfortunately, the quality of these sources is not always guaranteed. As additional complications, the data might be stored in a distributed way, or might even have to remain private. In this work, we address the question of how to learn robustly in such scenarios. Studying the problem through the lens of statistical learning theory, we derive a procedure that allows for learning from all available sources, yet automatically suppresses irrelevant or corrupted data. We show by extensive experiments that our method provides significant improvements over alternative approaches from robust statistics and distributed optimization. }, author = {Konstantinov, Nikola H and Lampert, Christoph}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, location = {Long Beach, CA, USA}, pages = {3488--3498}, publisher = {ML Research Press}, title = {{Robust learning from untrusted sources}}, volume = {97}, year = {2019}, } @article{6596, abstract = {It is well known that many problems in image recovery, signal processing, and machine learning can be modeled as finding zeros of the sum of maximal monotone and Lipschitz continuous monotone operators. Many papers have studied forward-backward splitting methods for finding zeros of the sum of two monotone operators in Hilbert spaces. Most of the proposed splitting methods in the literature have been proposed for the sum of maximal monotone and inverse-strongly monotone operators in Hilbert spaces. In this paper, we consider splitting methods for finding zeros of the sum of maximal monotone operators and Lipschitz continuous monotone operators in Banach spaces. We obtain weak and strong convergence results for the zeros of the sum of maximal monotone and Lipschitz continuous monotone operators in Banach spaces. Many already studied problems in the literature can be considered as special cases of this paper.}, author = {Shehu, Yekini}, issn = {1420-9012}, journal = {Results in Mathematics}, number = {4}, publisher = {Springer}, title = {{Convergence results of forward-backward algorithms for sum of monotone operators in Banach spaces}}, doi = {10.1007/s00025-019-1061-4}, volume = {74}, year = {2019}, } @article{6601, abstract = {There is increasing evidence that both mechanical and biochemical signals play important roles in development and disease. The development of complex organisms, in particular, has been proposed to rely on the feedback between mechanical and biochemical patterning events. This feedback occurs at the molecular level via mechanosensation but can also arise as an emergent property of the system at the cellular and tissue level. In recent years, dynamic changes in tissue geometry, flow, rheology, and cell fate specification have emerged as key platforms of mechanochemical feedback loops in multiple processes. Here, we review recent experimental and theoretical advances in understanding how these feedbacks function in development and disease.}, author = {Hannezo, Edouard B and Heisenberg, Carl-Philipp J}, issn = {00928674}, journal = {Cell}, number = {1}, pages = {12--25}, publisher = {Elsevier}, title = {{Mechanochemical feedback loops in development and disease}}, doi = {10.1016/j.cell.2019.05.052}, volume = {178}, year = {2019}, } @article{6607, abstract = {Acute myeloid leukemia (AML) is a heterogeneous disease with respect to its genetic and molecular basis and to patients´ outcome. Clinical, cytogenetic, and mutational data are used to classify patients into risk groups with different survival, however, within-group heterogeneity is still an issue. Here, we used a robust likelihood-based survival modeling approach and publicly available gene expression data to identify a minimal number of genes whose combined expression values were prognostic of overall survival. The resulting gene expression signature (4-GES) consisted of 4 genes (SOCS2, IL2RA, NPDC1, PHGDH), predicted patient survival as an independent prognostic parameter in several cohorts of AML patients (total, 1272 patients), and further refined prognostication based on the European Leukemia Net classification. An oncogenic role of the top scoring gene in this signature, SOCS2, was investigated using MLL-AF9 and Flt3-ITD/NPM1c driven mouse models of AML. SOCS2 promoted leukemogenesis as well as the abundance, quiescence, and activity of AML stem cells. Overall, the 4-GES represents a highly discriminating prognostic parameter in AML, whose clinical applicability is greatly enhanced by its small number of genes. The newly established role of SOCS2 in leukemia aggressiveness and stemness raises the possibility that the signature might even be exploitable therapeutically.}, author = {Nguyen, Chi Huu and Glüxam, Tobias and Schlerka, Angela and Bauer, Katharina and Grandits, Alexander M. and Hackl, Hubert and Dovey, Oliver and Zöchbauer-Müller, Sabine and Cooper, Jonathan L. and Vassiliou, George S. and Stoiber, Dagmar and Wieser, Rotraud and Heller, Gerwin}, journal = {Scientific Reports}, number = {1}, publisher = {Nature Publishing Group}, title = {{SOCS2 is part of a highly prognostic 4-gene signature in AML and promotes disease aggressiveness}}, doi = {10.1038/s41598-019-45579-0}, volume = {9}, year = {2019}, } @article{6608, abstract = {We use the canonical bases produced by the tri-partition algorithm in (Edelsbrunner and Ölsböck, 2018) to open and close holes in a polyhedral complex, K. In a concrete application, we consider the Delaunay mosaic of a finite set, we let K be an Alpha complex, and we use the persistence diagram of the distance function to guide the hole opening and closing operations. The dependences between the holes define a partial order on the cells in K that characterizes what can and what cannot be constructed using the operations. The relations in this partial order reveal structural information about the underlying filtration of complexes beyond what is expressed by the persistence diagram.}, author = {Edelsbrunner, Herbert and Ölsböck, Katharina}, journal = {Computer Aided Geometric Design}, pages = {1--15}, publisher = {Elsevier}, title = {{Holes and dependences in an ordered complex}}, doi = {10.1016/j.cagd.2019.06.003}, volume = {73}, year = {2019}, } @article{6609, abstract = {Mechanical systems facilitate the development of a hybrid quantum technology comprising electrical, optical, atomic and acoustic degrees of freedom1, and entanglement is essential to realize quantum-enabled devices. Continuous-variable entangled fields—known as Einstein–Podolsky–Rosen (EPR) states—are spatially separated two-mode squeezed states that can be used for quantum teleportation and quantum communication2. In the optical domain, EPR states are typically generated using nondegenerate optical amplifiers3, and at microwave frequencies Josephson circuits can serve as a nonlinear medium4,5,6. An outstanding goal is to deterministically generate and distribute entangled states with a mechanical oscillator, which requires a carefully arranged balance between excitation, cooling and dissipation in an ultralow noise environment. Here we observe stationary emission of path-entangled microwave radiation from a parametrically driven 30-micrometre-long silicon nanostring oscillator, squeezing the joint field operators of two thermal modes by 3.40 decibels below the vacuum level. The motion of this micromechanical system correlates up to 50 photons per second per hertz, giving rise to a quantum discord that is robust with respect to microwave noise7. Such generalized quantum correlations of separable states are important for quantum-enhanced detection8 and provide direct evidence of the non-classical nature of the mechanical oscillator without directly measuring its state9. This noninvasive measurement scheme allows to infer information about otherwise inaccessible objects, with potential implications for sensing, open-system dynamics and fundamental tests of quantum gravity. In the future, similar on-chip devices could be used to entangle subsystems on very different energy scales, such as microwave and optical photons.}, author = {Barzanjeh, Shabir and Redchenko, Elena and Peruzzo, Matilda and Wulf, Matthias and Lewis, Dylan and Arnold, Georg M and Fink, Johannes M}, journal = {Nature}, pages = {480--483}, publisher = {Nature Publishing Group}, title = {{Stationary entangled radiation from micromechanical motion}}, doi = {10.1038/s41586-019-1320-2}, volume = {570}, year = {2019}, } @article{151, abstract = {We construct planar bi-Sobolev mappings whose local volume distortion is bounded from below by a given function f∈Lp with p>1. More precisely, for any 1<q<(p+1)/2 we construct W1,q-bi-Sobolev maps with identity boundary conditions; for f∈L∞, we provide bi-Lipschitz maps. The basic building block of our construction are bi-Lipschitz maps which stretch a given compact subset of the unit square by a given factor while preserving the boundary. The construction of these stretching maps relies on a slight strengthening of the celebrated covering result of Alberti, Csörnyei, and Preiss for measurable planar sets in the case of compact sets. We apply our result to a model functional in nonlinear elasticity, the integrand of which features fast blowup as the Jacobian determinant of the deformation becomes small. For such functionals, the derivation of the equilibrium equations for minimizers requires an additional regularization of test functions, which our maps provide.}, author = {Fischer, Julian L and Kneuss, Olivier}, journal = {Journal of Differential Equations}, number = {1}, pages = {257 -- 311}, publisher = {Elsevier}, title = {{Bi-Sobolev solutions to the prescribed Jacobian inequality in the plane with L p data and applications to nonlinear elasticity}}, doi = {10.1016/j.jde.2018.07.045}, volume = {266}, year = {2019}, } @misc{13067, abstract = {Genetic incompatibilities contribute to reproductive isolation between many diverging populations, but it is still unclear to what extent they play a role if divergence happens with gene flow. In contact zones between the "Crab" and "Wave" ecotypes of the snail Littorina saxatilis divergent selection forms strong barriers to gene flow, while the role of postzygotic barriers due to selection against hybrids remains unclear. High embryo abortion rates in this species could indicate the presence of such barriers. Postzygotic barriers might include genetic incompatibilities (e.g. Dobzhansky-Muller incompatibilities) but also maladaptation, both expected to be most pronounced in contact zones. In addition, embryo abortion might reflect physiological stress on females and embryos independent of any genetic stress. We examined all embryos of >500 females sampled outside and inside contact zones of three populations in Sweden. Females' clutch size ranged from 0 to 1011 embryos (mean 130±123) and abortion rates varied between 0 and100% (mean 12%). We described female genotypes by using a hybrid index based on hundreds of SNPs differentiated between ecotypes with which we characterised female genotypes. We also calculated female SNP heterozygosity and inversion karyotype. Clutch size did not vary with female hybrid index and abortion rates were only weakly related to hybrid index in two sites but not at all in a third site. No additional variation in abortion rate was explained by female SNP heterozygosity, but increased female inversion heterozygosity added slightly to increased abortion. Our results show only weak and probably biologically insignificant postzygotic barriers contributing to ecotype divergence and the high and variable abortion rates were marginally, if at all, explained by hybrid index of females.}, author = {Johannesson, Kerstin and Zagrodzka, Zuzanna and Faria, Rui and Westram, Anja M and Butlin, Roger}, publisher = {Dryad}, title = {{Data from: Is embryo abortion a postzygotic barrier to gene flow between Littorina ecotypes?}}, doi = {10.5061/DRYAD.TB2RBNZWK}, 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}, } @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}, }