Search

Bridge functions, the missing link in the exact description of strong correlations, are indirectly extracted from specially designed molecular dynamics simulations of classical one-component plasma liquids and accurately parameterized. Their incorporation into an advanced integral equation theory description of Yukawa one-component plasma liquids and a novel dielectric formalism scheme for quantum one-component plasma liquids leads to an unprecedented agreement with available molecular dynamics simulations and new ab initio path integral Monte Carlo simulations, respectively., Comment: 7 pages, 5 figures, 2 tables, 27 pages of supplementary material

This paper introduces a new benchmark for large-scale image similarity detection. This benchmark is used for the Image Similarity Challenge at NeurIPS'21 (ISC2021). The goal is to determine whether a query image is a modified copy of any image in a reference corpus of size 1~million. The benchmark features a variety of image transformations such as automated transformations, hand-crafted image edits and machine-learning based manipulations. This mimics real-life cases appearing in social media, for example for integrity-related problems dealing with misinformation and objectionable content. The strength of the image manipulations, and therefore the difficulty of the benchmark, is calibrated according to the performance of a set of baseline approaches. Both the query and reference set contain a majority of "distractor" images that do not match, which corresponds to a real-life needle-in-haystack setting, and the evaluation metric reflects that. We expect the DISC21 benchmark to promote image copy detection as an important and challenging computer vision task and refresh the state of the art. Code and data are available at https://github.com/facebookresearch/isc2021

To extract the voice of a target speaker when mixed with a variety of other sounds, such as white and ambient noises or the voices of interfering speakers, we extend the Transformer network to attend the most relevant information with respect to the target speaker given the characteristics of his or her voices as a form of contextual information. The idea has a natural interpretation in terms of the selective attention theory. Specifically, we propose two models to incorporate the voice characteristics in Transformer based on different insights of where the feature selection should take place. Both models yield excellent performance, on par or better than published state-of-the-art models on the speaker extraction task, including separating speech of novel speakers not seen during training., Comment: 6 pages, 4 main figures, 2 supplemental figures

Incrementally training deep neural networks to recognize new classes is a challenging problem. Most existing class-incremental learning methods store data or use generative replay, both of which have drawbacks, while 'rehearsal-free' alternatives such as parameter regularization or bias-correction methods do not consistently achieve high performance. Here, we put forward a new strategy for class-incremental learning: generative classification. Rather than directly learning the conditional distribution p(y|x), our proposal is to learn the joint distribution p(x,y), factorized as p(x|y)p(y), and to perform classification using Bayes' rule. As a proof-of-principle, here we implement this strategy by training a variational autoencoder for each class to be learned and by using importance sampling to estimate the likelihoods p(x|y). This simple approach performs very well on a diverse set of continual learning benchmarks, outperforming generative replay and other existing baselines that do not store data., Comment: To appear in the IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPR-W) on Continual Learning in Computer Vision (CLVision) 2021

The thermodynamic and structural properties of two dimensional dense Yukawa liquids are studied with molecular dynamics simulations. The "exact" thermodynamic properties are simultaneously employed in an advanced scheme for the determination of an equation of state that shows an unprecedented level of accuracy for the internal energy, pressure and isothermal compressibility. The "exact" structural properties are utilized to formulate a novel empirical correction to the hypernetted-chain approach that leads to a very high accuracy level in terms of static correlations and thermodynamics., Comment: 13 pages, 7 figures, 13 page supplementary material

We prove sharp $L^1$ inequalities for the dyadic maximal function $M_T\phi$ when $\phi$ satisfies certain $L^1$ and $L^{\infty}$ conditions, Comment: 20 pages

Learning continually from non-stationary data streams is a long-standing goal and a challenging problem in machine learning. Recently, we have witnessed a renewed and fast-growing interest in continual learning, especially within the deep learning community. However, algorithmic solutions are often difficult to re-implement, evaluate and port across different settings, where even results on standard benchmarks are hard to reproduce. In this work, we propose Avalanche, an open-source end-to-end library for continual learning research based on PyTorch. Avalanche is designed to provide a shared and collaborative codebase for fast prototyping, training, and reproducible evaluation of continual learning algorithms., Comment: Official Website: https://avalanche.continualai.org

In moderately coupled Yukawa fluids, longitudinal mode dispersion is determined by the competition between kinetic and potential effects. In a recent paper [Khrapak and Cou\"edel, Phys. Rev. E 102, 033207 (2020)], a semi-phenomenological dispersion relation was constructed by the ad-hoc addition of the Bohm-Gross kinetic term to the generalized instantaneous excess bulk modulus, which showed very good agreement with simulations. In this paper, a nearly identical dispersion relation is derived in a rigorous manner based on a dielectric formulation with static local field corrections. At moderate coupling, this formalism is revealed to be more accurate than other successful theoretical approaches., Comment: 6 pages, 3 figures

The description of electronic exchange--correlation effects is of paramount importance for many applications in physics, chemistry, and beyond. In a recent Letter, Dornheim \textit{et al.} [\textit{Phys. Rev. Lett.}~\textbf{125}, 235001 (2020)] have presented the \emph{effective static approximation} (ESA) to the local field correction (LFC), which allows for the highly accurate estimation of electronic properties such as the interaction energy and the static structure factor. In the present work, we give an analytical parametrization of the LFC within ESA that is valid for any wave number, and available for the entire range of densities ($0.7\leq r_s \leq20$) and temperatures ($0\leq \theta\leq 4$) that are relevant for applications both in the ground state and in the warm dense matter regime. A short implementation in Python is provided, which can easily be incorporated into existing codes. In addition, we present an extensive analysis of the performance of ESA regarding the estimation of various quantities like the dynamic structure factor, static dielectric function, the electronically screened ion-potential, and also stopping power in electronic medium. In summary, we find that the ESA gives an excellent description of all these quantities in the warm dense matter regime, and only becomes inaccurate when the electrons start to form a strongly correlated electron liquid ($r_s\sim20$). Moreover, we note that the exact incorporation of exact asymptotic limits often leads to a superior accuracy compared to the neural-net representation of the static LFC [\textit{J.~Chem.~Phys.}~\textbf{151}, 194104 (2019)].

Despite impressive progress in deep learning, generalizing far beyond the training distribution is an important open challenge. In this work, we consider few-shot classification, and aim to shed light on what makes some novel classes easier to learn than others, and what types of learned representations generalize better. To this end, we define a new paradigm in terms of attributes -- simple building blocks of which concepts are formed -- as a means of quantifying the degree of relatedness of different concepts. Our empirical analysis reveals that supervised learning generalizes poorly to new attributes, but a combination of self-supervised pretraining with supervised finetuning leads to stronger generalization. The benefit of self-supervised pretraining and supervised finetuning is further investigated through controlled experiments using random splits of the attribute space, and we find that predictability of test attributes provides an informative estimate of a model's generalization ability., Comment: Technical report, 26 pages

In recent years, artificial neural networks have achieved state-of-the-art performance for predicting the responses of neurons in the visual cortex to natural stimuli. However, they require a time consuming parameter optimization process for accurately modeling the tuning function of newly observed neurons, which prohibits many applications including real-time, closed-loop experiments. We overcome this limitation by formulating the problem as $K$-shot prediction to directly infer a neuron's tuning function from a small set of stimulus-response pairs using a Neural Process. This required us to developed a Factorized Neural Process, which embeds the observed set into a latent space partitioned into the receptive field location and the tuning function properties. We show on simulated responses that the predictions and reconstructed receptive fields from the Factorized Neural Process approach ground truth with increasing number of trials. Critically, the latent representation that summarizes the tuning function of a neuron is inferred in a quick, single forward pass through the network. Finally, we validate this approach on real neural data from visual cortex and find that the predictive accuracy is comparable to -- and for small $K$ even greater than -- optimization based approaches, while being substantially faster. We believe this novel deep learning systems identification framework will facilitate better real-time integration of artificial neural network modeling into neuroscience experiments., Comment: 14 pages, 5 figures, NeurIPS 2020 conference paper

It has been conjectured that bridge functions remain nearly invariant along phase diagram lines of constant excess entropy for the class of R-simple liquids. In the companion paper, this hypothesis has been confirmed for Yukawa bridge functions outside the correlation void. In order to complete the testing of the invariance ansatz, Yukawa bridge functions are here computed inside the correlation void with the cavity distribution method and input from ultra-long molecular dynamics simulations featuring a tagged particle pair. A general methodology is developed for the design of the tagged pair interaction potential that leads to the acquisition of uniform statistics. An extrapolation technique is developed to determine the bridge function value at zero separation. The effect of different sources of errors is quantified. Yukawa bridge functions are demonstrated to be nearly isomorph invariant also in the short range. Generalization to all R-simple systems and practical implications are discussed., Comment: 19 pages, 10 figures, 3 tables, supplementary material

We propose an efficient method to learn deep local descriptors for instance-level recognition. The training only requires examples of positive and negative image pairs and is performed as metric learning of sum-pooled global image descriptors. At inference, the local descriptors are provided by the activations of internal components of the network. We demonstrate why such an approach learns local descriptors that work well for image similarity estimation with classical efficient match kernel methods. The experimental validation studies the trade-off between performance and memory requirements of the state-of-the-art image search approach based on match kernels. Compared to existing local descriptors, the proposed ones perform better in two instance-level recognition tasks and keep memory requirements lower. We experimentally show that global descriptors are not effective enough at large scale and that local descriptors are essential. We achieve state-of-the-art performance, in some cases even with a backbone network as small as ResNet18., Comment: ECCV 2020

Different advanced bridge function closures are utilized to investigate the structural and thermodynamic properties of dense Yukawa one-component plasma liquids within the framework of integral equation theory. The isomorph-based empirically modified hypernetted-chain, the variational modified hypernetted-chain, the Rogers-Young and the Ballone-Pastore-Galli-Gazzillo approaches are compared at the level of thermodynamic properties, radial distribution functions and bridge functions. The comparison, based on accuracy and computational speed, concludes that the two modified hypernetted-chain approaches are superior and singles out the isomorph-based variant as the most promising alternative to computer simulations of structural properties of dense Yukawa liquids. The possibility of further improvement through artificial cross-over to exact asymptotic limits is studied., Comment: 13 pages, 5 figures, 29 pages of supplementary material

The Lifshitz theory of van der Waals forces is utilized for the systematic calculation of the non-retarded room temperature Hamaker constants between 26 identical isotropic elemental metals that are embedded in vacuum or in pure water. The full spectral method, complemented with a Drude-like low frequency extrapolation, is employed for the elemental metals benefitting from the availability of extended-in-frequency reliable dielectric data. The simple spectral method is employed for pure water and three dielectric representations are explored. Numerical truncation and low frequency extrapolation effects are shown to be negligible. The accuracy of common Lifshitz approximations is quantified. The Hamaker constants for 100 metal combinations are reported; the geometric mixing rule is revealed to be highly accurate in vacuum and water., Comment: 13 pages; 2 figures; 7 tables

Despite impressive performance on numerous visual tasks, Convolutional Neural Networks (CNNs) --- unlike brains --- are often highly sensitive to small perturbations of their input, e.g. adversarial noise leading to erroneous decisions. We propose to regularize CNNs using large-scale neuroscience data to learn more robust neural features in terms of representational similarity. We presented natural images to mice and measured the responses of thousands of neurons from cortical visual areas. Next, we denoised the notoriously variable neural activity using strong predictive models trained on this large corpus of responses from the mouse visual system, and calculated the representational similarity for millions of pairs of images from the model's predictions. We then used the neural representation similarity to regularize CNNs trained on image classification by penalizing intermediate representations that deviated from neural ones. This preserved performance of baseline models when classifying images under standard benchmarks, while maintaining substantially higher performance compared to baseline or control models when classifying noisy images. Moreover, the models regularized with cortical representations also improved model robustness in terms of adversarial attacks. This demonstrates that regularizing with neural data can be an effective tool to create an inductive bias towards more robust inference., Comment: 14 pages, 7 figures, NeurIPS 2019

In this work we consider the problem of learning a classifier from noisy labels when a few clean labeled examples are given. The structure of clean and noisy data is modeled by a graph per class and Graph Convolutional Networks (GCN) are used to predict class relevance of noisy examples. For each class, the GCN is treated as a binary classifier, which learns to discriminate clean from noisy examples using a weighted binary cross-entropy loss function. The GCN-inferred "clean" probability is then exploited as a relevance measure. Each noisy example is weighted by its relevance when learning a classifier for the end task. We evaluate our method on an extended version of a few-shot learning problem, where the few clean examples of novel classes are supplemented with additional noisy data. Experimental results show that our GCN-based cleaning process significantly improves the classification accuracy over not cleaning the noisy data, as well as standard few-shot classification where only few clean examples are used.

To understand the function of cortical circuits it is necessary to classify their underlying cellular diversity. Traditional attempts based on comparing anatomical or physiological features of neurons and glia, while productive, have not resulted in a unified taxonomy of neural cell types. The recent development of single-cell transcriptomics has enabled, for the first time, systematic high-throughput profiling of large numbers of cortical cells and the generation of datasets that hold the promise of being complete, accurate and permanent. Statistical analyses of these data have revealed the existence of clear clusters, many of which correspond to cell types defined by traditional criteria, and which are conserved across cortical areas and species. To capitalize on these innovations and advance the field, we, the Copenhagen Convention Group, propose the community adopts a transcriptome-based taxonomy of the cell types in the adult mammalian neocortex. This core classification should be ontological, hierarchical and use a standardized nomenclature. It should be configured to flexibly incorporate new data from multiple approaches, developmental stages and a growing number of species, enabling improvement and revision of the classification. This community-based strategy could serve as a common foundation for future detailed analysis and reverse engineering of cortical circuits and serve as an example for cell type classification in other parts of the nervous system and other organs., Comment: 21 pages, 3 figures

Access to online visual search engines implies sharing of private user content - the query images. We introduce the concept of targeted mismatch attack for deep learning based retrieval systems to generate an adversarial image to conceal the query image. The generated image looks nothing like the user intended query, but leads to identical or very similar retrieval results. Transferring attacks to fully unseen networks is challenging. We show successful attacks to partially unknown systems, by designing various loss functions for the adversarial image construction. These include loss functions, for example, for unknown global pooling operation or unknown input resolution by the retrieval system. We evaluate the attacks on standard retrieval benchmarks and compare the results retrieved with the original and adversarial image.

Standard artificial neural networks suffer from the well-known issue of catastrophic forgetting, making continual or lifelong learning difficult for machine learning. In recent years, numerous methods have been proposed for continual learning, but due to differences in evaluation protocols it is difficult to directly compare their performance. To enable more structured comparisons, we describe three continual learning scenarios based on whether at test time task identity is provided and--in case it is not--whether it must be inferred. Any sequence of well-defined tasks can be performed according to each scenario. Using the split and permuted MNIST task protocols, for each scenario we carry out an extensive comparison of recently proposed continual learning methods. We demonstrate substantial differences between the three scenarios in terms of difficulty and in terms of how efficient different methods are. In particular, when task identity must be inferred (i.e., class incremental learning), we find that regularization-based approaches (e.g., elastic weight consolidation) fail and that replaying representations of previous experiences seems required for solving this scenario., Comment: Extended version of work presented at the NeurIPS Continual Learning workshop (2018); 18 pages, 5 figures, 6 tables. Related to arXiv:1809.10635

We propose a kernelized deep local-patch descriptor based on efficient match kernels of neural network activations. Response of each receptive field is encoded together with its spatial location using explicit feature maps. Two location parametrizations, Cartesian and polar, are used to provide robustness to a different types of canonical patch misalignment. Additionally, we analyze how the conventional architecture, i.e. a fully connected layer attached after the convolutional part, encodes responses in a spatially variant way. In contrary, explicit spatial encoding is used in our descriptor, whose potential applications are not limited to local-patches. We evaluate the descriptor on standard benchmarks. Both versions, encoding 32x32 or 64x64 patches, consistently outperform all other methods on all benchmarks. The number of parameters of the model is independent of the input patch resolution.

In numerous realizations of complex plasmas, dust-dust interactions are characterized by two screening lengths and are thus better described by a combination of Yukawa potentials. The present work investigates the static correlations and the thermodynamics of repulsive dense bi-Yukawa fluids based on the fact that such strongly coupled systems exhibit isomorph invariance. The strong virial-potential energy correlations are demonstrated with the aid of molecular dynamics simulations, an accurate analytical expression for the isomorph family of curves is obtained and an empirical expression for the fluid-solid phase-coexistence line is proposed. The isomorph-based empirically modified hypernetted-chain approach, grounded on the ansatz of isomorph invariant bridge functions, is then extended to such systems and the resulting structural properties show an excellent agreement with the results of computer simulations. A simple and accurate closed-form expression is obtained for the excess internal energy of dense bi-Yukawa fluids by capitalizing on the compact parameterization offered by the Rosenfeld-Tarazona decomposition in combination with the Rosenfeld scaling, which opens up the energy route to thermodynamics., Comment: 11 pages, 6 figures

Semi-supervised learning is becoming increasingly important because it can combine data carefully labeled by humans with abundant unlabeled data to train deep neural networks. Classic methods on semi-supervised learning that have focused on transductive learning have not been fully exploited in the inductive framework followed by modern deep learning. The same holds for the manifold assumption---that similar examples should get the same prediction. In this work, we employ a transductive label propagation method that is based on the manifold assumption to make predictions on the entire dataset and use these predictions to generate pseudo-labels for the unlabeled data and train a deep neural network. At the core of the transductive method lies a nearest neighbor graph of the dataset that we create based on the embeddings of the same network.Therefore our learning process iterates between these two steps. We improve performance on several datasets especially in the few labels regime and show that our work is complementary to current state of the art., Comment: Accepted to CVPR 2019

Isomorph theory is employed in order to establish a mapping between the bridge function of Coulomb and Yukawa one-component plasmas. Within an exact invariance ansatz for the bridge functions and by capitalizing on the availability of simulation-extracted Coulomb bridge functions, an analytical Yukawa bridge function is derived which is inserted into the integral theory framework. In spite of its simplicity and computational speed, the proposed integral approach exhibits an excellent agreement with computer simulations of dense Yukawa liquids without invoking adjustable parameters., Comment: 12 pages, 6 figures, extensive tabulations in the 38-page supplemental material

We propose a multiple-kernel local-patch descriptor based on efficient match kernels from pixel gradients. It combines two parametrizations of gradient position and direction, each parametrization provides robustness to a different type of patch mis-registration: polar parametrization for noise in the patch dominant orientation detection, Cartesian for imprecise location of the feature point. Combined with whitening of the descriptor space, that is learned with or without supervision, the performance is significantly improved. We analyze the effect of the whitening on patch similarity and demonstrate its semantic meaning. Our unsupervised variant is the best performing descriptor constructed without the need of labeled data. Despite the simplicity of the proposed descriptor, it competes well with deep learning approaches on a number of different tasks.

A major obstacle to developing artificial intelligence applications capable of true lifelong learning is that artificial neural networks quickly or catastrophically forget previously learned tasks when trained on a new one. Numerous methods for alleviating catastrophic forgetting are currently being proposed, but differences in evaluation protocols make it difficult to directly compare their performance. To enable more meaningful comparisons, here we identified three distinct scenarios for continual learning based on whether task identity is known and, if it is not, whether it needs to be inferred. Performing the split and permuted MNIST task protocols according to each of these scenarios, we found that regularization-based approaches (e.g., elastic weight consolidation) failed when task identity needed to be inferred. In contrast, generative replay combined with distillation (i.e., using class probabilities as "soft targets") achieved superior performance in all three scenarios. Addressing the issue of efficiency, we reduced the computational cost of generative replay by integrating the generative model into the main model by equipping it with generative feedback or backward connections. This Replay-through-Feedback approach substantially shortened training time with no or negligible loss in performance. We believe this to be an important first step towards making the powerful technique of generative replay scalable to real-world continual learning applications., Comment: 17 pages, 8 figures, 4 tables

Classical models describe primary visual cortex (V1) as a filter bank of orientation-selective linear-nonlinear (LN) or energy models, but these models fail to predict neural responses to natural stimuli accurately. Recent work shows that models based on convolutional neural networks (CNNs) lead to much more accurate predictions, but it remains unclear which features are extracted by V1 neurons beyond orientation selectivity and phase invariance. Here we work towards systematically studying V1 computations by categorizing neurons into groups that perform similar computations. We present a framework to identify common features independent of individual neurons' orientation selectivity by using a rotation-equivariant convolutional neural network, which automatically extracts every feature at multiple different orientations. We fit this model to responses of a population of 6000 neurons to natural images recorded in mouse primary visual cortex using two-photon imaging. We show that our rotation-equivariant network not only outperforms a regular CNN with the same number of feature maps, but also reveals a number of common features shared by many V1 neurons, which deviate from the typical textbook idea of V1 as a bank of Gabor filters. Our findings are a first step towards a powerful new tool to study the nonlinear computations in V1.

Recent advances in deep learning have brought to the fore models that can make multiple computational steps in the service of completing a task; these are capable of describ- ing long-term dependencies in sequential data. Novel recurrent attention models over possibly large external memory modules constitute the core mechanisms that enable these capabilities. Our work addresses learning subtler and more complex underlying temporal dynamics in language modeling tasks that deal with sparse sequential data. To this end, we improve upon these recent advances, by adopting concepts from the field of Bayesian statistics, namely variational inference. Our proposed approach consists in treating the network parameters as latent variables with a prior distribution imposed over them. Our statistical assumptions go beyond the standard practice of postulating Gaussian priors. Indeed, to allow for handling outliers, which are prevalent in long observed sequences of multivariate data, multivariate t-exponential distributions are imposed. On this basis, we proceed to infer corresponding posteriors; these can be used for inference and prediction at test time, in a way that accounts for the uncertainty in the available sparse training data. Specifically, to allow for our approach to best exploit the merits of the t-exponential family, our method considers a new t-divergence measure, which generalizes the concept of the Kullback-Leibler divergence. We perform an extensive experimental evaluation of our approach, using challenging language modeling benchmarks, and illustrate its superiority over existing state-of-the-art techniques.

The relational data model offers unrivaled rigor and precision in defining data structure and querying complex data. Yet the use of relational databases in scientific data pipelines is limited due to their perceived unwieldiness. We propose a simplified and conceptually refined relational data model named DataJoint. The model includes a language for schema definition, a language for data queries, and diagramming notation for visualizing entities and relationships among them. The model adheres to the principle of entity normalization, which requires that all data -- both stored and derived -- must be represented by well-formed entity sets. DataJoint's data query language is an algebra on entity sets with five operators that provide matching capabilities to those of other relational query languages with greater clarity due to entity normalization. Practical implementations of DataJoint have been adopted in neuroscience labs for fluent interaction with scientific data pipelines., Comment: 26 pages, 5 figures

State of the art image retrieval performance is achieved with CNN features and manifold ranking using a k-NN similarity graph that is pre-computed off-line. The two most successful existing approaches are temporal filtering, where manifold ranking amounts to solving a sparse linear system online, and spectral filtering, where eigen-decomposition of the adjacency matrix is performed off-line and then manifold ranking amounts to dot-product search online. The former suffers from expensive queries and the latter from significant space overhead. Here we introduce a novel, theoretically well-founded hybrid filtering approach allowing full control of the space-time trade-off between these two extremes. Experimentally, we verify that our hybrid method delivers results on par with the state of the art, with lower memory demands compared to spectral filtering approaches and faster compared to temporal filtering.

Calcium imaging has revolutionized systems neuroscience, providing the ability to image large neural populations with single-cell resolution. The resulting datasets are quite large, which has presented a barrier to routine open sharing of this data, slowing progress in reproducible research. State of the art methods for analyzing this data are based on non-negative matrix factorization (NMF); these approaches solve a non-convex optimization problem, and are effective when good initializations are available, but can break down in low-SNR settings where common initialization approaches fail. Here we introduce an approach to compressing and denoising functional imaging data. The method is based on a spatially-localized penalized matrix decomposition (PMD) of the data to separate (low-dimensional) signal from (temporally-uncorrelated) noise. This approach can be applied in parallel on local spatial patches and is therefore highly scalable, does not impose non-negativity constraints or require stringent identifiability assumptions (leading to significantly more robust results compared to NMF), and estimates all parameters directly from the data, so no hand-tuning is required. We have applied the method to a wide range of functional imaging data (including one-photon, two-photon, three-photon, widefield, somatic, axonal, dendritic, calcium, and voltage imaging datasets): in all cases, we observe ~2-4x increases in SNR and compression rates of 20-300x with minimal visible loss of signal, with no adjustment of hyperparameters; this in turn facilitates the process of demixing the observed activity into contributions from individual neurons. We focus on two challenging applications: dendritic calcium imaging data and voltage imaging data in the context of optogenetic stimulation. In both cases, we show that our new approach leads to faster and much more robust extraction of activity from the data., Comment: 36 pages, 18 figures

Neural attention (NA) has become a key component of sequence-to-sequence models that yield state-of-the-art performance in as hard tasks as abstractive document summarization (ADS) and video captioning (VC). NA mechanisms perform inference of context vectors; these constitute weighted sums of deterministic input sequence encodings, adaptively sourced over long temporal horizons. Inspired from recent work in the field of amortized variational inference (AVI), in this work we consider treating the context vectors generated by soft-attention (SA) models as latent variables, with approximate finite mixture model posteriors inferred via AVI. We posit that this formulation may yield stronger generalization capacity, in line with the outcomes of existing applications of AVI to deep networks. To illustrate our method, we implement it and experimentally evaluate it considering challenging ADS, VC, and MT benchmarks. This way, we exhibit its improved effectiveness over state-of-the-art alternatives., Comment: Submitted for Review

In this work we present a novel unsupervised framework for hard training example mining. The only input to the method is a collection of images relevant to the target application and a meaningful initial representation, provided e.g. by pre-trained CNN. Positive examples are distant points on a single manifold, while negative examples are nearby points on different manifolds. Both types of examples are revealed by disagreements between Euclidean and manifold similarities. The discovered examples can be used in training with any discriminative loss. The method is applied to unsupervised fine-tuning of pre-trained networks for fine-grained classification and particular object retrieval. Our models are on par or are outperforming prior models that are fully or partially supervised.

In this paper we address issues with image retrieval benchmarking on standard and popular Oxford 5k and Paris 6k datasets. In particular, annotation errors, the size of the dataset, and the level of challenge are addressed: new annotation for both datasets is created with an extra attention to the reliability of the ground truth. Three new protocols of varying difficulty are introduced. The protocols allow fair comparison between different methods, including those using a dataset pre-processing stage. For each dataset, 15 new challenging queries are introduced. Finally, a new set of 1M hard, semi-automatically cleaned distractors is selected. An extensive comparison of the state-of-the-art methods is performed on the new benchmark. Different types of methods are evaluated, ranging from local-feature-based to modern CNN based methods. The best results are achieved by taking the best of the two worlds. Most importantly, image retrieval appears far from being solved., Comment: CVPR 2018

For every Banach space $X$ with a Schauder basis consider the Banach algebra $\mathbb{R} I\oplus\mathcal{K}_\mathrm{diag}(X)$ of all diagonal operators that are of the form $\lambda I + K$. We prove that $\mathbb{R} I\oplus\mathcal{K}_\mathrm{diag}(X)$ is a Calkin algbra i.e., there exists a Banach space $\mathcal{Y}_X$ so that the Calkin algebra of $\mathcal{Y}_X$ is isomorphic as a Banach algebra to $\mathbb{R} I\oplus\mathcal{K}_\mathrm{diag}(X)$. Among other applications of this theorem we obtain that certain hereditarily indecomposable spaces and the James spaces $J_p$ and their duals endowed with natural multiplications are Calkin algebras, that all non-reflexive Banach spaces with unconditional bases are isomorphic as Banach spaces to Calkin algebras, and that sums of reflexive spaces with unconditional bases with certain James-Tsirelson type spaces are isomorphic as Banach spaces to Calkin algebras., Comment: 49 pages

Image descriptors based on activations of Convolutional Neural Networks (CNNs) have become dominant in image retrieval due to their discriminative power, compactness of representation, and search efficiency. Training of CNNs, either from scratch or fine-tuning, requires a large amount of annotated data, where a high quality of annotation is often crucial. In this work, we propose to fine-tune CNNs for image retrieval on a large collection of unordered images in a fully automated manner. Reconstructed 3D models obtained by the state-of-the-art retrieval and structure-from-motion methods guide the selection of the training data. We show that both hard-positive and hard-negative examples, selected by exploiting the geometry and the camera positions available from the 3D models, enhance the performance of particular-object retrieval. CNN descriptor whitening discriminatively learned from the same training data outperforms commonly used PCA whitening. We propose a novel trainable Generalized-Mean (GeM) pooling layer that generalizes max and average pooling and show that it boosts retrieval performance. Applying the proposed method to the VGG network achieves state-of-the-art performance on the standard benchmarks: Oxford Buildings, Paris, and Holidays datasets., Comment: TPAMI 2018. arXiv admin note: substantial text overlap with arXiv:1604.02426

Severe background clutter is challenging in many computer vision tasks, including large-scale image retrieval. Global descriptors, that are popular due to their memory and search efficiency, are especially prone to corruption by such a clutter. Eliminating the impact of the clutter on the image descriptor increases the chance of retrieving relevant images and prevents topic drift due to actually retrieving the clutter in the case of query expansion. In this work, we propose a novel salient region detection method. It captures, in an unsupervised manner, patterns that are both discriminative and common in the dataset. Saliency is based on a centrality measure of a nearest neighbor graph constructed from regional CNN representations of dataset images. The descriptors derived from the salient regions improve particular object retrieval, most noticeably in a large collections containing small objects.

We cast shape matching as metric learning with convolutional networks. We break the end-to-end process of image representation into two parts. Firstly, well established efficient methods are chosen to turn the images into edge maps. Secondly, the network is trained with edge maps of landmark images, which are automatically obtained by a structure-from-motion pipeline. The learned representation is evaluated on a range of different tasks, providing improvements on challenging cases of domain generalization, generic sketch-based image retrieval or its fine-grained counterpart. In contrast to other methods that learn a different model per task, object category, or domain, we use the same network throughout all our experiments, achieving state-of-the-art results in multiple benchmarks., Comment: ECCV 2018

We propose a multiple-kernel local-patch descriptor based on efficient match kernels of patch gradients. It combines two parametrizations of gradient position and direction, each parametrization provides robustness to a different type of patch miss-registration: polar parametrization for noise in the patch dominant orientation detection, Cartesian for imprecise location of the feature point. Even though handcrafted, the proposed method consistently outperforms the state-of-the-art methods on two local patch benchmarks., Comment: To appear in the British Machine Vision Conference (BMVC), September 2017

Location recognition is commonly treated as visual instance retrieval on "street view" imagery. The dataset items and queries are panoramic views, i.e. groups of images taken at a single location. This work introduces a novel panorama-to-panorama matching process, either by aggregating features of individual images in a group or by explicitly constructing a larger panorama. In either case, multiple views are used as queries. We reach near perfect location recognition on a standard benchmark with only four query views.