Filters








2,928 Hits in 4.2 sec

Escaping Saddle Points Efficiently with Occupation-Time-Adapted Perturbations [article]

Xin Guo, Jiequn Han, Mahan Tajrobehkar, Wenpin Tang
2022 arXiv   pre-print
After integrating this mechanism into the framework of perturbed gradient descent (PGD) and perturbed accelerated gradient descent (PAGD), two new algorithms are proposed: perturbed gradient descent adapted  ...  In this mechanism, perturbations are adapted to the history of states via the notion of occupation time.  ...  Perturbed Gradient Descent Adapted with Occupation Time PGD adds a uniform random perturbation when stuck at saddle points.  ... 
arXiv:2005.04507v3 fatcat:dciocsvkkfahdastijayxcfqcm

Functional Path Optimisation for Exploration in Continuous Occupancy Maps [article]

Gilad Francis, Lionel Ott, Fabio Ramos
2018 arXiv   pre-print
Autonomous exploration is a complex task where the robot moves through an unknown environment with the goal of mapping it.  ...  In this work, we formulate exploration as a variational problem which allows us to directly optimise in the space of trajectories using functional gradient methods, searching for the Next Best Path (NBP  ...  In addition, stochastic gradient descent (SGD) is used to enable real-time performance.  ... 
arXiv:1805.01079v1 fatcat:h7yw2ij225fs7nbl62koa3breq

Avoiding Occupancy Detection from Smart Meter using Adversarial Machine Learning [article]

ibrahim Yilmaz, Ambareen Siraj
2020 arXiv   pre-print
Detecting the occupancy of a home is straightforward with time of use information as there is a strong correlation between occupancy and electricity usage.  ...  Essentially, the proposed privacy-preserving framework is designed to mask real-time or near real-time electricity usage information using calculated optimum noise without compromising users' billing systems  ...  After adding subtle perturbation by gradient descent, the same model predicts the same data incorrectly as occupied.  ... 
arXiv:2010.12640v1 fatcat:4yum3vamjzcqzhfxgakn47dd3y

A Privacy-Preserving Energy Consumption Scheme for Smart Meters with Adversarial Machine Learning

Ibrahim Yilmaz, Ambareen Siraj
2021 IEEE Access  
Detecting the occupancy of a home is straightforward with time of use information as there is a strong correlation between occupancy and electricity usage.  ...  With the proposed AMLODA approach working to protect consumers' privacy, occupancy detection attacks are demonstrated to be mitigated with the MCC values of the attack models converging to zero with no  ...  After adding subtle perturbation by gradient descent, the same model predicts the same data incorrectly as occupied.  ... 
doi:10.1109/access.2021.3057525 fatcat:kekugdgaofgxzfepefy6vqu25i

A Full Quantum Eigensolver for Quantum Chemistry Simulations [article]

Shijie Wei, Hang Li, GuiLu Long
2020 arXiv   pre-print
The gradient descent iteration depth has a favorable complexity that is logarithmically dependent on the system size and inverse of the precision.  ...  Here, we propose a full quantum eigensolver (FQE) algorithm to calculate the molecular ground energies and electronic structures using quantum gradient descent.  ...  For very noisy situations that do not allow many iterations, FQE can be combined with perturbation theory that give the ground state and its energy in chemical precision with one time iteration.  ... 
arXiv:1908.07927v2 fatcat:243y3rgtanbgpotxkbu36xycym

A Full Quantum Eigensolver for Quantum Chemistry Simulations

Shijie Wei, Hang Li, GuiLu Long
2020 Research  
The gradient descent iteration depth has a favorable complexity that is logarithmically dependent on the system size and inverse of the precision.  ...  Here, we propose a full quantum eigensolver (FQE) algorithm to calculate the molecular ground energies and electronic structures using quantum gradient descent.  ...  Here, we present an approximate method to find the ground state and its energy by using the gradient descent algorithm and perturbation theory.  ... 
doi:10.34133/2020/1486935 pmid:32274468 pmcid:PMC7125455 fatcat:vj3khb4agnfgpduqt3sdjlmy3e

Stochastic Functional Gradient Path Planning in Occupancy Maps [article]

Gilad Francis, Lionel Ott, Fabio Ramos
2017 arXiv   pre-print
We show that the stochasticity of the samples is crucial for the planner and present comparisons to other state-of-the-art planning methods in both simulation and with real occupancy data.  ...  In this work, we improve our previous work on stochastic functional gradient planners.  ...  The sampled gradient g(t i ) can be viewed as a path perturbation g(t i ) = ∇ ξ U(ξ n )(t i )Υ(t i ).  ... 
arXiv:1705.05987v1 fatcat:m5durd3bejgmzksrwf656yn24a

Hilbert maps: scalable continuous occupancy mapping with stochastic gradient descent

Fabio Ramos, Lionel Ott
2015 Robotics: Science and Systems XI  
We devise a new technique for environment representation through continuous occupancy mapping that improves on the popular occupancy grip maps in two fundamental aspects: 1) it does not assume an a priori  ...  We present results with three types of approximations, Random Fourier, Nyström and a novel sparse projection.  ...  Convergence of stochastic gradient descent In this experiment we compare the convergence of SGD for the full batch case where all the data is presented to the algorithm multiple times and the incremental  ... 
doi:10.15607/rss.2015.xi.002 dblp:conf/rss/RamosO15 fatcat:gbr6jnohivhyvbmsfejqy43xjy

Hilbert maps: Scalable continuous occupancy mapping with stochastic gradient descent

Fabio Ramos, Lionel Ott
2016 The international journal of robotics research  
We devise a new technique for environment representation through continuous occupancy mapping that improves on the popular occupancy grip maps in two fundamental aspects: 1) it does not assume an a priori  ...  We present results with three types of approximations, Random Fourier, Nyström and a novel sparse projection.  ...  Convergence of stochastic gradient descent In this experiment we compare the convergence of SGD for the full batch case where all the data is presented to the algorithm multiple times and the incremental  ... 
doi:10.1177/0278364916684382 fatcat:vd3kfmficvddrmc2ynkk5srbuu

Dynamics-aware Adversarial Attack of 3D Sparse Convolution Network [article]

An Tao and Yueqi Duan and He Wang and Ziyi Wu and Pengliang Ji and Haowen Sun and Jie Zhou and Jiwen Lu
2021 arXiv   pre-print
To address this issue, we propose a Leaded Gradient Method (LGM) and show the significant effects of the lagged gradient.  ...  Compared with the dynamic-unaware methods, LGM achieves about 20% lower mIoU averagely on the ScanNet and S3DIS datasets. LGM also outperforms the recent point cloud attacks.  ...  architecture at time t + 1.  ... 
arXiv:2112.09428v1 fatcat:qz4u6vrm5zfstd6z6u6yeqgh6u

A Generic and Model-Agnostic Exemplar Synthetization Framework for Explainable AI [article]

Antonio Barbalau, Adrian Cosma, Radu Tudor Ionescu, Marius Popescu
2020 arXiv   pre-print
Moreover, we compare our framework (available at https://github.com/antoniobarbalau/exemplar) with a model-dependent approach based on gradient descent, proving that our framework obtains equally-good  ...  exemplars in a shorter computational time.  ...  Therefore, we experimented using gradient descent with momentum.  ... 
arXiv:2006.03896v3 fatcat:3y6q4tb7xrglll65izcn6nwluy

Active Exploration and Mapping via Iterative Covariance Regulation over Continuous SE(3) Trajectories [article]

Shumon Koga and Arash Asgharivaskasi and Nikolay Atanasov
2021 arXiv   pre-print
We introduce a differentiable field of view formulation, and derive iCR via the gradient descent method to iteratively update an open-loop control sequence in continuous space so that the covariance of  ...  We demonstrate autonomous exploration and uncertainty reduction in simulated occupancy grid environments.  ...  While in [8] the gradient descent is applied via perturbation method to obtain an approximated gradient, our approach develops an explicit gradient formulation.  ... 
arXiv:2103.05819v1 fatcat:q6vnjh4dnffy3o3pbtjak43nyq

Attack and defence in cellular decision-making: lessons from machine learning [article]

Thomas J. Rademaker, Emmanuel Bengio, Paul François
2019 arXiv   pre-print
We then apply a gradient-descent approach from machine learning to different cellular decision-making models, and we reveal the existence of two regimes characterized by the presence or absence of a critical  ...  Machine learning algorithms are sensitive to meaningless (or "adversarial") perturbations.  ...  First, the latter perturbations are difficult to find through gradient descent (as illustrated by the many steps to reach the boundary in Fig. 3 A) .  ... 
arXiv:1807.04270v2 fatcat:sclnxpvfubayvns4u6u7ui2spe

Experimental Realization of a Quantum Autoencoder: The Compression of Qutrits via Machine Learning

Alex Pepper, Nora Tischler, Geoff J. Pryde
2019 Physical Review Letters  
We also show that the device is able to perform with minimal prior information about the quantum data or physical system and is robust to perturbations during its optimization routine.  ...  With quantum resources a precious commodity, their efficient use is highly desirable. Quantum autoencoders have been proposed as a way to reduce quantum memory requirements.  ...  We probe the junk output, calculate the cost function, and use a classical gradient descent optimization routine to adjust the unitary [17] .  ... 
doi:10.1103/physrevlett.122.060501 fatcat:qm4bcsnz4je2tah6ses3svesca

Fooling the classifier: Ligand antagonism and adversarial examples [article]

Thomas J Rademaker, Emmanuel Bengio, Paul Francois
2018 bioRxiv   pre-print
We then use a gradient-descent approach to compare different adaptive proofreading models, and we reveal the existence of two qualitatively different regimes characterized by the presence or absence of  ...  Machine learning algorithms are sensitive to so-called adversarial perturbations.  ...  First, the latter perturbations are difficult to find through gradient descent (as illustrated by the many steps to reach the boundary in Fig. 3 A) .  ... 
doi:10.1101/366724 fatcat:qal66hxrwvcxvmjx6v3jb6xhuq
« Previous Showing results 1 — 15 out of 2,928 results