Showing 1–13 of 13 results for author: Krylov, D

Moonwalk: Inverse-Forward Differentiation

Authors: Dmitrii Krylov, Armin Karamzade, Roy Fox

Abstract: Backpropagation, while effective for gradient computation, falls short in addressing memory consumption, limiting scalability. This work explores forward-mode gradient computation as an alternative in invertible networks, showing its potential to reduce the memory footprint without substantial drawbacks. We introduce a novel technique based on a vector-inverse-Jacobian product that accelerates the… ▽ More Backpropagation, while effective for gradient computation, falls short in addressing memory consumption, limiting scalability. This work explores forward-mode gradient computation as an alternative in invertible networks, showing its potential to reduce the memory footprint without substantial drawbacks. We introduce a novel technique based on a vector-inverse-Jacobian product that accelerates the computation of forward gradients while retaining the advantages of memory reduction and preserving the fidelity of true gradients. Our method, Moonwalk, has a time complexity linear in the depth of the network, unlike the quadratic time complexity of naïve forward, and empirically reduces computation time by several orders of magnitude without allocating more memory. We further accelerate Moonwalk by combining it with reverse-mode differentiation to achieve time complexity comparable with backpropagation while maintaining a much smaller memory footprint. Finally, we showcase the robustness of our method across several architecture choices. Moonwalk is the first forward-based method to compute true gradients in invertible networks in computation time comparable to backpropagation and using significantly less memory. △ Less

Submitted 21 February, 2024; originally announced February 2024.

Align Your Intents: Offline Imitation Learning via Optimal Transport

Authors: Maksim Bobrin, Nazar Buzun, Dmitrii Krylov, Dmitry V. Dylov

Abstract: Offline reinforcement learning (RL) addresses the problem of sequential decision-making by learning optimal policy through pre-collected data, without interacting with the environment. As yet, it has remained somewhat impractical, because one rarely knows the reward explicitly and it is hard to distill it retrospectively. Here, we show that an imitating agent can still learn the desired behavior m… ▽ More Offline reinforcement learning (RL) addresses the problem of sequential decision-making by learning optimal policy through pre-collected data, without interacting with the environment. As yet, it has remained somewhat impractical, because one rarely knows the reward explicitly and it is hard to distill it retrospectively. Here, we show that an imitating agent can still learn the desired behavior merely from observing the expert, despite the absence of explicit rewards or action labels. In our method, AILOT (Aligned Imitation Learning via Optimal Transport), we involve special representation of states in a form of intents that incorporate pairwise spatial distances within the data. Given such representations, we define intrinsic reward function via optimal transport distance between the expert's and the agent's trajectories. We report that AILOT outperforms state-of-the art offline imitation learning algorithms on D4RL benchmarks and improves the performance of other offline RL algorithms in the sparse-reward tasks. △ Less

Submitted 20 February, 2024; originally announced February 2024.

Interpreting X-ray absorption spectra of Vanadyl Phthalocyanines Spin Qubit Candidates using a Machine Learning-Assisted Approach

Authors: Jaehyun Lee, Corina M. Urdaniz, Stefano Reale, Kyungju Noh, Denis Krylov, Andrin Doll, Luciano Colazzo, Yujeong Bae, Christoph Wolf, Fabio Donati

Abstract: The magnetic dilution of Vanadyl phthalocyanine (VOPc) within the isostructural diamagnetic Titanyl phthalocyanine (TiOPc) affords promising molecular spin qubit platforms for solid-state quantum computing. The development of quantitative methods for determining how the interactions with a supporting substrate impact the electronic structure of the system are fundamental to determine their potenti… ▽ More The magnetic dilution of Vanadyl phthalocyanine (VOPc) within the isostructural diamagnetic Titanyl phthalocyanine (TiOPc) affords promising molecular spin qubit platforms for solid-state quantum computing. The development of quantitative methods for determining how the interactions with a supporting substrate impact the electronic structure of the system are fundamental to determine their potential integration in physical devices. In this work we propose a combined approach based on X-ray absorption spectroscopy (XAS), atomic multiplet calculations, and density functional theory (DFT) to investigate the 3d orbital level structure of VOPc on TiOPc/Ag(100). We characterize VOPc in different molecular environments realized by changing the thickness of TiOPc interlayer and adsorption configuration on Ag(100). Depending on the molecular film structure, we find characteristic XAS features that we analyze using atomic multiplet calculations. We use a Bayesian optimization algorithm to accelerate the parameter search process in the multiplet calculations and identify the ground state properties, such as the 3d orbital occupancy and splitting, as well as intra-atomic interactions. Our analysis indicates that VOPc retains its spin S = 1/2 character in all configurations. Conversely, the energy separation and sequence of the unoccupied V 3d orbitals sensitively depend on the interaction with the surface and TiOPc interlayer. We validate the atomic orbital picture obtained from the multiplet model by comparison with DFT, which further allows us to understand the VOPc electronic properties using a molecular orbital description. △ Less

Submitted 22 March, 2024; v1 submitted 28 September, 2023; originally announced September 2023.

Comments: https://journals.aps.org/prb/accepted/fa07aO95X4f10b4334a769f8107f50ef629b9b212

Learning to Design Analog Circuits to Meet Threshold Specifications

Authors: Dmitrii Krylov, Pooya Khajeh, Junhan Ouyang, Thomas Reeves, Tongkai Liu, Hiba Ajmal, Hamidreza Aghasi, Roy Fox

Abstract: Automated design of analog and radio-frequency circuits using supervised or reinforcement learning from simulation data has recently been studied as an alternative to manual expert design. It is straightforward for a design agent to learn an inverse function from desired performance metrics to circuit parameters. However, it is more common for a user to have threshold performance criteria rather t… ▽ More Automated design of analog and radio-frequency circuits using supervised or reinforcement learning from simulation data has recently been studied as an alternative to manual expert design. It is straightforward for a design agent to learn an inverse function from desired performance metrics to circuit parameters. However, it is more common for a user to have threshold performance criteria rather than an exact target vector of feasible performance measures. In this work, we propose a method for generating from simulation data a dataset on which a system can be trained via supervised learning to design circuits to meet threshold specifications. We moreover perform the to-date most extensive evaluation of automated analog circuit design, including experimenting in a significantly more diverse set of circuits than in prior work, covering linear, nonlinear, and autonomous circuit configurations, and show that our method consistently reaches success rate better than 90% at 5% error margin, while also improving data efficiency by upward of an order of magnitude. A demo of this system is available at circuits.streamlit.app △ Less

Submitted 25 July, 2023; originally announced July 2023.

Comments: in proceedings of ICML 23

Development of a Scanning Tunneling Microscope for Variable Temperature Electron Spin Resonance

Authors: Jiyoon Hwang, Denis Krylov, Robertus J. G. Elbertse, Sangwon Yoon, Taehong Ahn, Jeongmin Oh, Lei Fang, Won-jun Jang, Franklin H. Cho, Andreas J. Heinrich, Yujeong Bae

Abstract: Recent advances in increasing the spectroscopic energy resolution in scanning tunneling microscopy (STM) have been achieved by integrating electron spin resonance (ESR) with STM. Here, we demonstrate the design and performance of a home-built STM capable of ESR at temperatures ranging from 1 K to 10 K. The STM is incorporated with a home-built Joule-Thomson refrigerator and a 2-axis vector magnet.… ▽ More Recent advances in increasing the spectroscopic energy resolution in scanning tunneling microscopy (STM) have been achieved by integrating electron spin resonance (ESR) with STM. Here, we demonstrate the design and performance of a home-built STM capable of ESR at temperatures ranging from 1 K to 10 K. The STM is incorporated with a home-built Joule-Thomson refrigerator and a 2-axis vector magnet. Our STM design allows for the deposition of atoms and molecules directly into the cold STM, eliminating the need to extract the sample for deposition. In addition, we adopt two methods to apply radio-frequency (RF) voltages to the tunnel junction, the early design of wiring to the STM tip directly, and a more recent idea to use an RF antenna. Direct comparisons of ESR results measured using the two methods and simulations of electric field distribution around the tunnel junction show that, despite their different designs and capacitive couplings to the tunnel junction, there is no discernible difference in the driving and detection of ESR. Furthermore, at a magnetic field of 1.6 T, we observe ESR signals (near 40 GHz) sustained up to 10 K, which is the highest temperature for ESR-STM measurement reported to date, to the best of our knowledge. Although the ESR intensity exponentially decreases with increasing temperature, our ESR-STM system with low noise at the tunnel junction allows us to measure weak ESR signals with intensities in the sub-fA range. Our new design of ESR-STM, which is operational in a large frequency and temperature range, can broaden the use of ESR spectroscopy in STM and enable the simple modification of existing STM systems, which will hopefully accelerate a generalized use of ESR-STM. △ Less

Submitted 23 August, 2022; originally announced August 2022.

Spin Resonance Amplitude and Frequency of a Single Atom on a Surface in a Vector Magnetic Field

Authors: **kyung Kim, Won-jun Jang, Thi Hong Bui, Deung-jang Choi, Christoph Wolf, Fernando Delgado, Denis Krylov, Soonhyeong Lee, Sangwon Yoon, Christopher P. Lutz, Andreas J. Heinrich, Yujeong Bae

Abstract: We used electron spin resonance (ESR) combined with scanning tunneling microscopy (STM) to measure hydrogenated Ti (spin-1/2) atoms at low-symmetry binding sites on MgO in vector magnetic fields. We found strongly anisotropic g-values in all three spatial directions. Interestingly, the amplitude and lineshape of the ESR signals are also strongly dependent on the angle of the field. We conclude tha… ▽ More We used electron spin resonance (ESR) combined with scanning tunneling microscopy (STM) to measure hydrogenated Ti (spin-1/2) atoms at low-symmetry binding sites on MgO in vector magnetic fields. We found strongly anisotropic g-values in all three spatial directions. Interestingly, the amplitude and lineshape of the ESR signals are also strongly dependent on the angle of the field. We conclude that the Ti spin is aligned along the magnetic field, while the tip spin follows its strong magnetic anisotropy. Our results show the interplay between the tip and surface spins in determining the ESR signals and highlight the precision of ESR-STM to identify the single atom's spin states. △ Less

Submitted 17 March, 2021; originally announced March 2021.

Orbital-resolved single atom magnetism measured with X-ray absorption spectroscopy

Authors: Aparajita Singha, Daria Sostina, Christoph Wolf, Safa L. Ahmed, Denis Krylov, Luciano Colazzo, Pierluigi Gargiani, Stefano Agrestini, Woo-Suk Noh, Jae-Hoon Park, Marina Pivetta, Stefano Rusponi, Harald Brune, Andreas J. Heinrich, Alessandro Barla, Fabio Donati

Abstract: Lanthanide atoms and molecules are promising candidates for atomic data storage and quantum logic due to the long magnetic lifetime of their electron quantum states. Accessing these states through electrical transport requires the engineering of their electronic configuration down to the level of individual atomic orbitals. Here, we address the magnetism of surface-supported lanthanide atoms, clus… ▽ More Lanthanide atoms and molecules are promising candidates for atomic data storage and quantum logic due to the long magnetic lifetime of their electron quantum states. Accessing these states through electrical transport requires the engineering of their electronic configuration down to the level of individual atomic orbitals. Here, we address the magnetism of surface-supported lanthanide atoms, clusters, and films with orbital selectivity using X-ray absorption spectroscopy and magnetic circular dichroism. We exploit the selection rules of electric dipole transitions to reveal the occupation and magnetism of the valence electrons of Gd and Ho deposited on MgO/Ag(100). Comparing our results with multiplet calculations and density functional theory, we identify a charge transfer mechanism that leaves the lanthanide species in an unconventional singly ionized configuration. Our approach allows determining the role of valence electrons on the quantum level structure of lanthanide-based nanostructures. △ Less

Submitted 20 December, 2020; originally announced December 2020.

Fast XML/HTML for Haskell: XML TypeLift

Authors: Michał J. Gajda, Dmitry Krylov

Abstract: The paper presents and compares a range of parsers with and without data map** for conversion between XML and Haskell. The best performing parser competes favorably with the fastest tools available in other languages and is, thus, suitable for use in large-scale data analysis. The best performing parser also allows software developers of intermediate-level Haskell programming skills to start pro… ▽ More The paper presents and compares a range of parsers with and without data map** for conversion between XML and Haskell. The best performing parser competes favorably with the fastest tools available in other languages and is, thus, suitable for use in large-scale data analysis. The best performing parser also allows software developers of intermediate-level Haskell programming skills to start processing large numbers of XML documents soon after finding the relevant XML Schema from a simple internet search, without the need for specialist prior knowledge or skills. We hope that this unique combination of parser performance and usability will provide a new standard for XML map** to high-level languages. △ Less

Submitted 5 November, 2020; originally announced November 2020.

Reinforcement Learning Framework for Deep Brain Stimulation Study

Authors: Dmitrii Krylov, Remi Tachet, Romain Laroche, Michael Rosenblum, Dmitry V. Dylov

Abstract: Malfunctioning neurons in the brain sometimes operate synchronously, reportedly causing many neurological diseases, e.g. Parkinson's. Suppression and control of this collective synchronous activity are therefore of great importance for neuroscience, and can only rely on limited engineering trials due to the need to experiment with live human brains. We present the first Reinforcement Learning gym… ▽ More Malfunctioning neurons in the brain sometimes operate synchronously, reportedly causing many neurological diseases, e.g. Parkinson's. Suppression and control of this collective synchronous activity are therefore of great importance for neuroscience, and can only rely on limited engineering trials due to the need to experiment with live human brains. We present the first Reinforcement Learning gym framework that emulates this collective behavior of neurons and allows us to find suppression parameters for the environment of synthetic degenerate models of neurons. We successfully suppress synchrony via RL for three pathological signaling regimes, characterize the framework's stability to noise, and further remove the unwanted oscillations by engaging multiple PPO agents. △ Less

Submitted 22 February, 2020; originally announced February 2020.

Comments: 7 pages + 1 references, 7 figures. arXiv admin note: text overlap with arXiv:1909.12154

Journal ref: IJCAI 2020, pp. 2847-2854

Reinforcement learning for suppression of collective activity in oscillatory ensembles

Authors: Dmitriy Krylov, Dmitry V. Dylov, Michael Rosenblum

Abstract: We present a use of modern data-based machine learning approaches to suppress self-sustained collective oscillations typically signaled by ensembles of degenerative neurons in the brain. The proposed hybrid model relies on two major components: an environment of oscillators and a policy-based reinforcement learning block. We report a model-agnostic synchrony control based on proximal policy optimi… ▽ More We present a use of modern data-based machine learning approaches to suppress self-sustained collective oscillations typically signaled by ensembles of degenerative neurons in the brain. The proposed hybrid model relies on two major components: an environment of oscillators and a policy-based reinforcement learning block. We report a model-agnostic synchrony control based on proximal policy optimization and two artificial neural networks in an Actor-Critic configuration. A class of physically meaningful reward functions enabling the suppression of collective oscillatory mode is proposed. The synchrony suppression is demonstrated for two models of neuronal populations -- for the ensembles of globally coupled limit-cycle Bonhoeffer-van der Pol oscillators and for the bursting Hindmarsh--Rose neurons. △ Less

Submitted 16 January, 2020; v1 submitted 25 September, 2019; originally announced September 2019.

Comments: 8 pages, 8 figures

Collimation of a Circulating Beam in the U_70 Synchrotron by Use of Reflections in Axially - Oriented Crystals

Authors: A. G. Afonin, V. T. Baranov, M. K. Bulgakov, I. S. Voinov, V. N. Gorlov, I. V. Ivanova, D. M. Krylov, A. N. Lunn'kov, V. A. Maisheev, S. F. Reshetnikov, D. A. Savin, E. A. Syshchikov, V. I. Terekhov, Yu. A. Chesnokov, P. N. Chirkov, I. A. Yazynin

Abstract: The possibilities of the extraction and collimation of a circulating beam by a new method due to the reflection of particles in crystals with axial orientation were experimentally investigated in the Fall-2010 run at the U_70 synchrotron. Such crystals have positive features, because the axial potential is five times larger than the planar potential. It has been shown that the collimation efficien… ▽ More The possibilities of the extraction and collimation of a circulating beam by a new method due to the reflection of particles in crystals with axial orientation were experimentally investigated in the Fall-2010 run at the U_70 synchrotron. Such crystals have positive features, because the axial potential is five times larger than the planar potential. It has been shown that the collimation efficiency can reach 90% due to axial effects in the crystal. Losses of the circulating beam on a collimator have been reduced by several times; this makes it possible to suppress the muon jet near the steel collimator of the circulating beam. △ Less

Submitted 28 March, 2011; originally announced March 2011.

Comments: 6 pages, 7 figures

Journal ref: JETP Lett.93:187-190,2011

Crystal-Assisted Beam Extraction and Collimation at the U-70 Circular Accelerator

Authors: A. G. Afonin, V. T. Baranov, S. Bellucci, S. A. Belov, S. Bini, V. N. Gorlov, G. Giannini, A. D. Ermolaev, I. V. Ivanova, D. M. Krylov, V. A. Maisheev, D. A. Savin, E. A. Syshchikov, V. I. Terekhov, V. N. Chepegin, Yu. A. Chesnokov, P. N. Chirkov, I. A. Yazynin

Abstract: New crystal technique - array of bent strips and a fan-type reflector, based on thin straight plates - have been used for research of extraction and collimation a circulating beam in the U-70 accelerator at the energy 50 GeV and 1.3 GeV. It is shown, that new devices can effectively steer a beam in a wide energy range. For protons with energy 50 GeV efficiency of extraction and collimation about 9… ▽ More New crystal technique - array of bent strips and a fan-type reflector, based on thin straight plates - have been used for research of extraction and collimation a circulating beam in the U-70 accelerator at the energy 50 GeV and 1.3 GeV. It is shown, that new devices can effectively steer a beam in a wide energy range. For protons with energy 50 GeV efficiency of extraction and collimation about 90 % has been achieved which is record for this method. Reduction of particle losses in 2-3 times was observed also in accelerator at application of different crystals in comparison with the usual one-stage collimation scheme of beam with a steel absorber. △ Less

Submitted 14 February, 2011; originally announced February 2011.

Comments: 14 pages, 12 figures

Report number: IHEP preprint 2010-12

Journal ref: Instruments and Experimental Techniques, Vol.54, 1, 2011

Parity and Relative Parity in Knot Theory

Authors: D. Yu. Krylov, V. O. Manturov

Abstract: In the present paper we give a simple proof of the fact that the set of virtual links with orientable atoms is closed. More precisely, the theorem states that if two virtual diagrams $K$ and $K'$ have orientable atoms and they are equivalent by Reidemeister moves, then there is a sequence of diagrams $K = K_1 \to...\to K_n=K'$ all having orientable atoms where $K_i$ is obtained from $K_{i-1}$ by a… ▽ More In the present paper we give a simple proof of the fact that the set of virtual links with orientable atoms is closed. More precisely, the theorem states that if two virtual diagrams $K$ and $K'$ have orientable atoms and they are equivalent by Reidemeister moves, then there is a sequence of diagrams $K = K_1 \to...\to K_n=K'$ all having orientable atoms where $K_i$ is obtained from $K_{i-1}$ by a Reidemeister move. The initial proof heavily relies on the topology of virtual links and was published in \cite{IM}. Our proof is based on the notion of parity which was introduced by the second named author in 2009. We split the set of crossings of a virtual link diagram into sets of {\it odd} and {\it even} in accordance with a fixed rule. The rule must only satisfy several conditions of Reidemeister's type. Then one can construct functorial map**s of link diagrams by using parity. The concept of parity allows one to introduce new invariants and strengthen well-known ones \cite{Ma1}. △ Less

Submitted 30 December, 2010; originally announced January 2011.

Comments: 9 pages, 4 figures