arXiv subjects

Alexey Potapov

Publications and source records attributed to Alexey Potapov.

At least 37 records · Page 2Linked to original sources

arXiv2020

GoodPoint: unsupervised learning of keypoint detection and description

This paper introduces a new algorithm for unsupervised learning of keypoint detectors and descriptors, which demonstrates fast convergence and good performance across different datasets. The training procedure uses homographic transformation of images. The proposed model learns to detect points and generate descriptors on pairs of transformed images, which are easy for it to distinguish and repeatedly detect. The trained model follows SuperPoint architecture for ease of comparison, and demonstrates similar performance on natural images from HPatches dataset, and better performance on retina images from Fundus Image Registration Dataset, which contain low number of corner-like features. For HPatches and other datasets, coverage was also computed to provide better estimation of model quality.

cs.CV↗

arXiv2020

Ice coverage of dust grains in cold astrophysical environments

Surface processes on cosmic solids in cold astrophysical environments lead to gas phase depletion and molecular complexity. Most astrophysical models assume that the molecular ice forms a thick multilayer substrate, not interacting with the dust surface. In contrast, we present experimental results demonstrating the importance of the surface for porous grains. We show that cosmic dust grains may be covered by a few monolayers of ice only. This implies that the role of dust surface structure, composition, and reactivity in models describing surface processes in cold interstellar, protostellar, and protoplanetary environments has to be reevaluated.

astro-ph.GA↗

arXiv2020

Thermal formation of ammonium carbamate on the surface of laboratory analogues of carbonaceous grains in protostellar envelopes and planet-forming disks

The catalytic role of dust grain surfaces in the thermal reaction CO2 + 2NH3 $\rightarrow$ NH4+NH2COO was recently demonstrated by our group. The rate coefficients for the reaction at 80 K on the surface of nanometre-sized carbon and silicate grains were measured to be up to three times higher compared to the reaction rate coefficients measured on KBr. In this study, the reaction was performed on carbon grains and on KBr in the extended temperature range of 50 - 80 K and with the addition of water ice. The reaction activation energy was found to be about 3 times lower on grains compared to the corresponding ice layer on KBr. Thus, the catalytic role of the dust grain surface in the studied reaction can be related to a reduction of the reaction barrier. Addition of water to NH3:CO2 ice on grains slowed the reaction down. At the H2O:CO2 ratio of 5:1, the reaction was not detected on the experimental timescale. This result calls into question the thermal formation of ammonium carbamate in dense molecular clouds and outer regions of protostellar and protoplanetary environments with dominating water ice mantle chemistry. However, it can still happen in inner regions of protostellar and protoplanetary environments in crystalline ices.

astro-ph.GA↗

arXiv2019

High-resolution gas phase spectroscopy of molecules desorbed from an ice surface: a proof-of-principle study

High-resolution gas phase spectroscopy techniques in the microwave, millimeter-wave and terahertz spectral ranges can be used to study complex organic molecules desorbed from interstellar ice analogues surface with a high sensitivity. High-resolution gas phase spectroscopy gives unambiguous information about the molecular composition, the molecular structure, and transition frequencies needed for their detection by radio telescopes in various interstellar and circumstellar environments. The results will be useful not only for interpreting astronomical spectra and understanding astrophysical processes, but also for more general studies of gas-surface chemistry. This paper presents a new experimental approach based on a combination of a chirped-pulse Fourier transform microwave spectrometer detection and a low temperature surface desorption experiment. The experimental set-up is benchmarked on the desorption of ammonia ice detected by high-resolution gas phase microwave spectroscopy.

astro-ph.IM↗

arXiv2019

Constraining the detectability of water ice in debris disks

Water ice is important for the evolution and preservation of life. Identifying the distribution of water ice in debris disks is therefore of great interest in the field of astrobiology. Furthermore, icy dust grains are expected to play important roles throughout the entire planet formation process. However, currently available observations only allow deriving weak conclusions about the existence of water ice in debris disks. We investigate whether it is feasible to detect water ice in typical debris disk systems. We take the following ice destruction mechanisms into account: sublimation of ice, dust production through planetesimal collisions, and photosputtering by UV-bright central stars. We consider icy dust mixture particles with various shapes consisting of amorphous ice, crystalline ice, astrosilicate, and vacuum inclusions. We calculated optical properties of inhomogeneous icy dust mixtures using effective medium theories, that is, Maxwell-Garnett rules. Subsequently, we generated synthetic debris disk observables, such as spectral energy distributions and spatially resolved thermal reemission and scattered light intensity and polarization maps with our code DMS. We find that the prominent $\sim$ 3 $μ\rm{m}$ and 44 $μ\rm{m}$ water ice features can be potentially detected in future observations of debris disks with the James Webb Space Telescope and the Space Infrared telescope for Cosmology and Astrophysics. We show that the sublimation of ice, collisions between planetesimals, and photosputtering caused by UV sources clearly affect the observational appearance of debris disk systems. In addition, highly porous ice tends to produce highly polarized radiation at around 3 $μ\rm{m}$. Finally, the location of the ice survival line is determined by various dust properties such as a fractional ratio of ice versus dust, physical states of ice, and the porosity of icy grains.

astro-ph.EP↗

arXiv2019

Differentiable Probabilistic Logic Networks

Probabilistic logic reasoning is a central component of such cognitive architectures as OpenCog. However, as an integrative architecture, OpenCog facilitates cognitive synergy via hybridization of different inference methods. In this paper, we introduce a differentiable version of Probabilistic Logic networks, which rules operate over tensor truth values in such a way that a chain of reasoning steps constructs a computation graph over tensors that accepts truth values of premises from the knowledge base as input and produces truth values of conclusions as output. This allows for both learning truth values of premises and formulas for rules (specified in a form with trainable weights) by backpropagation combining subsymbolic optimization and symbolic reasoning.

cs.AI↗

arXiv2019

Photodesorption of water ice from dust grains and thermal desorption of cometary ices studied by the INSIDE experiment

A new experimental set-up INterStellar Ice-Dust Experiment (INSIDE), was designed for studying cosmic grain analogues represented by ice-coated carbon- and silicate-based dust grains. In the new instrument, we can simulate physical and chemical conditions prevailing in interstellar and circumstellar environments. The set-up combines ultrahigh vacuum and low temperature conditions with infrared spectroscopy and mass spectrometry. Using INSIDE, we plan to investigate physical and chemical processes, such as adsorption, desorption, and formation of molecules, on the surface of dust/ice samples. First experiments on the photodesorption of water ice molecules from the surface of silicate and carbon grains by UV photons revealed a strong influence of the surface properties on the desorption yield, in particular in the monolayer regime. In the second experiment, the thermal desorption of cometary ice analogues composed of six molecular components was studied for the first time. Co-desorption of CO2 and CH3OH with O2 indicates that at high O2 concentrations in cometary or interstellar ices "heavy" ice molecules can be partly trapped in O2 and release to the gas phase much earlier than expected. This effect could explain astronomical detections of complex organic molecules in cold dense interstellar clouds.

astro-ph.IM↗

arXiv2019

Evidence of surface catalytic effect on cosmic dust grain analogues: the ammonia and carbon dioxide surface reaction

Surface chemistry on cosmic dust grains plays an important role in the formation of molecules at low temperatures in the interstellar and circumstellar environments. For the first time, we experimentally put in evidence the catalytic role of dust surfaces using the thermal reaction CO2 + 2NH3 = NH4+NH2COO-, which is also a proxy of radical-radical reactions. Nanometre-sized amorphous silicate and carbon grains produced in our laboratory were used as grain analogues. Surface catalysis on grains accelerates the kinetics of the reaction studied at a temperature of 80 K by a factor of up to 3 compared to the reaction occurring in the molecular solid. The evidence of the catalytic effect of grain surfaces opens a door for experiments and calculations on the surface formation of interstellar and circumstellar molecules on dust. Ammonium carbamate on the surface of grains or released intact into protostellar or protoplanetary disk phases can give start to a network of prebiotic reactions. Therefore, there should be a great interest to search for ammonium carbamate and its daughter molecule, carbamic acid, in interstellar clouds, protostellar envelopes, and protoplanetary disks.

astro-ph.GA↗

arXiv2018

Temperature programed desorption of water ice from the surface of amorphous carbon and silicate grains as related to planet-forming disks

Understanding the history and evolution of small bodies, such as dust grains and comets, in planet-forming disks is very important to reveal the architectural laws responsible for the creation of planetary systems. These small bodies in cold regions of the disks are typically considered as mixtures of dust particles with molecular ices, where ices cover the surface of a dust core or are actually physically mixed with dust. Whilst the first case, ice-on-dust, has been intensively studied in the laboratory in recent decades, the second case, ice-mixed-with-dust, present uncharted territory. This work is the first laboratory study of the temperature-programmed desorption (TPD) of water ice mixed with amorphous carbon and silicate grains. We show that the kinetics of desorption of H2O ice depends strongly on the dust/ice mass ratio, probably, due to the desorption of water molecules from a large surface of fractal clusters composed of carbon or silicate grains. In addition, it is shown that water ice molecules are differently bound to silicate grains in contrast to carbon. The results provide a link between the structure and morphology of small cosmic bodies and the kinetics of desorption of water ice included in them.

astro-ph.GA↗

arXiv2018

Improving Deep Models of Person Re-identification for Cross-Dataset Usage

Person re-identification (Re-ID) is the task of matching humans across cameras with non-overlapping views that has important applications in visual surveillance. Like other computer vision tasks, this task has gained much with the utilization of deep learning methods. However, existing solutions based on deep learning are usually trained and tested on samples taken from same datasets, while in practice one need to deploy Re-ID systems for new sets of cameras for which labeled data is unavailable. Here, we mitigate this problem for one state-of-the-art model, namely, metric embedding trained with the use of the triplet loss function, although our results can be extended to other models. The contribution of our work consists in developing a method of training the model on multiple datasets, and a method for its online practically unsupervised fine-tuning. These methods yield up to 19.1% improvement in Rank-1 score in the cross-dataset evaluation.

cs.CV↗

arXiv2018

Metric Embedding Autoencoders for Unsupervised Cross-Dataset Transfer Learning

Cross-dataset transfer learning is an important problem in person re-identification (Re-ID). Unfortunately, not too many deep transfer Re-ID models exist for realistic settings of practical Re-ID systems. We propose a purely deep transfer Re-ID model consisting of a deep convolutional neural network and an autoencoder. The latent code is divided into metric embedding and nuisance variables. We then utilize an unsupervised training method that does not rely on co-training with non-deep models. Our experiments show improvements over both the baseline and competitors' transfer learning models.

cs.CV↗

arXiv2018

Genetic algorithms with DNN-based trainable crossover as an example of partial specialization of general search

Universal induction relies on some general search procedure that is doomed to be inefficient. One possibility to achieve both generality and efficiency is to specialize this procedure w.r.t. any given narrow task. However, complete specialization that implies direct mapping from the task parameters to solutions (discriminative models) without search is not always possible. In this paper, partial specialization of general search is considered in the form of genetic algorithms (GAs) with a specialized crossover operator. We perform a feasibility study of this idea implementing such an operator in the form of a deep feedforward neural network. GAs with trainable crossover operators are compared with the result of complete specialization, which is also represented as a deep neural network. Experimental results show that specialized GAs can be more efficient than both general GAs and discriminative models.

cs.NE↗

arXiv2018

HyperNets and their application to learning spatial transformations

In this paper we propose a conceptual framework for higher-order artificial neural networks. The idea of higher-order networks arises naturally when a model is required to learn some group of transformations, every element of which is well-approximated by a traditional feedforward network. Thus the group as a whole can be represented as a hyper network. One of typical examples of such groups is spatial transformations. We show that the proposed framework, which we call HyperNets, is able to deal with at least two basic spatial transformations of images: rotation and affine transformation. We show that HyperNets are able not only to generalize rotation and affine transformation, but also to compensate the rotation of images bringing them into canonical forms.

cs.NE↗

arXiv2018

Vision System for AGI: Problems and Directions

What frameworks and architectures are necessary to create a vision system for AGI? In this paper, we propose a formal model that states the task of perception within AGI. We show the role of discriminative and generative models in achieving efficient and general solution of this task, thus specifying the task in more detail. We discuss some existing generative and discriminative models and demonstrate their insufficiency for our purposes. Finally, we discuss some architectural dilemmas and open questions.

cs.CV↗

arXiv2018

Semantic Image Retrieval by Uniting Deep Neural Networks and Cognitive Architectures

Image and video retrieval by their semantic content has been an important and challenging task for years, because it ultimately requires bridging the symbolic/subsymbolic gap. Recent successes in deep learning enabled detection of objects belonging to many classes greatly outperforming traditional computer vision techniques. However, deep learning solutions capable of executing retrieval queries are still not available. We propose a hybrid solution consisting of a deep neural network for object detection and a cognitive architecture for query execution. Specifically, we use YOLOv2 and OpenCog. Queries allowing the retrieval of video frames containing objects of specified classes and specified spatial arrangement are implemented.

cs.IR↗

arXiv2018

Low temperature optical properties of interstellar and circumstellar icy silicate grain analogues in the mid-infrared spectral region

Two different silicate/water ice mixtures representing laboratory analogues of interstellar and circumstellar icy grains were produced in the laboratory. For the first time, optical constants, the real and imaginary parts of the complex refractive index, of such silicate/water ice mixtures were experimentally determined in the mid-infrared spectral region at low temperatures. In addition, optical constants of pure water ice and pure silicates were derived in the laboratory. Two sets of constants were compared, namely, measured constants calculated from the transmission spectra of silicate/water ice samples and effective constants calculated from the optical constants of pure silicates and pure water ice samples using different mixing rules (effective medium approaches). Differences between measured and effective constants show that a mixing (averaging) of optical constants of water ice and silicates for the determination of the optical properties of silicate/ice mixtures can lead to incorrect results. Also, it is shown that a part of water ice molecules is trapped in/on silicate grains and does not desorb up to 200 K. Our unique data are just in time with respect to the new challenging space mission, James Webb Space Telescope, which will be able to bring novel detailed information on interstellar and circumstellar grains, and suitable laboratory data are extremely important for the decoding of astronomical spectra.

astro-ph.GA↗

arXiv2016

A Step from Probabilistic Programming to Cognitive Architectures

Probabilistic programming is considered as a framework, in which basic components of cognitive architectures can be represented in unified and elegant fashion. At the same time, necessity of adopting some component of cognitive architectures for extending capabilities of probabilistic programming languages is pointed out. In particular, implicit specification of generative models via declaration of concepts and links between them is proposed, and usefulness of declarative knowledge for achieving efficient inference is briefly discussed.

cs.AI↗

arXiv2013

Universal Empathy and Ethical Bias for Artificial General Intelligence

Rational agents are usually built to maximize rewards. However, AGI agents can find undesirable ways of maximizing any prior reward function. Therefore value learning is crucial for safe AGI. We assume that generalized states of the world are valuable - not rewards themselves, and propose an extension of AIXI, in which rewards are used only to bootstrap hierarchical value learning. The modified AIXI agent is considered in the multi-agent environment, where other agents can be either humans or other "mature" agents, which values should be revealed and adopted by the "infant" AGI agent. General framework for designing such empathic agent with ethical bias is proposed also as an extension of the universal intelligence model. Moreover, we perform experiments in the simple Markov environment, which demonstrate feasibility of our approach to value learning in safe AGI.

cs.AI↗