arXiv subjects

Ravi Kumar

Publications and source records attributed to Ravi Kumar.

At least 127 records · Page 7Linked to original sources

arXiv2022

Parsimonious Learning-Augmented Caching

Learning-augmented algorithms -- in which, traditional algorithms are augmented with machine-learned predictions -- have emerged as a framework to go beyond worst-case analysis. The overarching goal is to design algorithms that perform near-optimally when the predictions are accurate yet retain certain worst-case guarantees irrespective of the accuracy of the predictions. This framework has been successfully applied to online problems such as caching where the predictions can be used to alleviate uncertainties. In this paper we introduce and study the setting in which the learning-augmented algorithm can utilize the predictions parsimoniously. We consider the caching problem -- which has been extensively studied in the learning-augmented setting -- and show that one can achieve quantitatively similar results but only using a sublinear number of predictions.

cs.DS↗

arXiv2022

Geometric quenches in quasi-disordered lattice system

While global quantum quench has been extensively used in the literature to understand the localization-delocalization transition for the one-dimensional quantum spin chain, the effect of geometric quench (which corresponds to a sudden change of the geometry of the chain) in the context of such transitions is yet to be well understood. In this work, we investigate the effect of geometric quench in the Aubry-Andre model, which supports localization-delocalization transition even in one dimension. We study the spreading of the entanglement and the site-occupation with time and find many interesting features that can be used to characterize localization-delocalization transition. We observe that geometric quench causes a power-law type growth of the entanglement entropy in the delocalized phase in contrast to the linear growth which is found in the global quench studies. Remarkably, we also find that the saturation values in the Many-body localized (MBL) phase obey Area law in contrast to the usual volume law which is a signature feature of the MBL phase in the context of global quench.

cond-mat.dis-nn↗

arXiv2021

Private Rank Aggregation in Central and Local Models

In social choice theory, (Kemeny) rank aggregation is a well-studied problem where the goal is to combine rankings from multiple voters into a single ranking on the same set of items. Since rankings can reveal preferences of voters (which a voter might like to keep private), it is important to aggregate preferences in such a way to preserve privacy. In this work, we present differentially private algorithms for rank aggregation in the pure and approximate settings along with distribution-independent utility upper and lower bounds. In addition to bounds in the central model, we also present utility bounds for the local model of differential privacy.

cs.DS↗

arXiv2021

User-Level Private Learning via Correlated Sampling

Most works in learning with differential privacy (DP) have focused on the setting where each user has a single sample. In this work, we consider the setting where each user holds $m$ samples and the privacy protection is enforced at the level of each user's data. We show that, in this setting, we may learn with a much fewer number of users. Specifically, we show that, as long as each user receives sufficiently many samples, we can learn any privately learnable class via an $(ε, δ)$-DP algorithm using only $O(\log(1/δ)/ε)$ users. For $ε$-DP algorithms, we show that we can learn using only $O_ε(d)$ users even in the local model, where $d$ is the probabilistic representation dimension. In both cases, we show a nearly-matching lower bound on the number of users required. A crucial component of our results is a generalization of global stability [Bun et al., FOCS 2020] that allows the use of public randomness. Under this relaxed notion, we employ a correlated sampling strategy to show that the global stability can be boosted to be arbitrarily close to one, at a polynomial expense in the number of samples.

cs.LG↗

arXiv2021

Learning-Augmented Weighted Paging

We consider a natural semi-online model for weighted paging, where at any time the algorithm is given predictions, possibly with errors, about the next arrival of each page. The model is inspired by Belady's classic optimal offline algorithm for unweighted paging, and extends the recently studied model for learning-augmented paging (Lykouris and Vassilvitskii, 2018) to the weighted setting. For the case of perfect predictions, we provide an $\ell$-competitive deterministic and an $O(\log \ell)$-competitive randomized algorithm, where $\ell$ is the number of distinct weight classes. Both these bounds are tight, and imply an $O(\log W)$- and $O(\log \log W)$-competitive ratio, respectively, when the page weights lie between $1$ and $W$. Previously, it was not known how to use these predictions in the weighted setting and only bounds of $k$ and $O(\log k)$ were known, where $k$ is the cache size. Our results also generalize to the interleaved paging setting and to the case of imperfect predictions, with the competitive ratios degrading smoothly from $O(\ell)$ and $O(\log \ell)$ to $O(k)$ and $O(\log k)$, respectively, as the prediction error increases. Our results are based on several insights on structural properties of Belady's algorithm and the sequence of page arrival predictions, and novel potential functions that incorporate these predictions. For the case of unweighted paging, the results imply a very simple potential function based proof of the optimality of Belady's algorithm, which may be of independent interest.

cs.DS↗

arXiv2021

Logarithmic Regret from Sublinear Hints

We consider the online linear optimization problem, where at every step the algorithm plays a point $x_t$ in the unit ball, and suffers loss $\langle c_t, x_t\rangle$ for some cost vector $c_t$ that is then revealed to the algorithm. Recent work showed that if an algorithm receives a hint $h_t$ that has non-trivial correlation with $c_t$ before it plays $x_t$, then it can achieve a regret guarantee of $O(\log T)$, improving on the bound of $Θ(\sqrt{T})$ in the standard setting. In this work, we study the question of whether an algorithm really requires a hint at every time step. Somewhat surprisingly, we show that an algorithm can obtain $O(\log T)$ regret with just $O(\sqrt{T})$ hints under a natural query model; in contrast, we also show that $o(\sqrt{T})$ hints cannot guarantee better than $Ω(\sqrt{T})$ regret. We give two applications of our result, to the well-studied setting of optimistic regret bounds and to the problem of online learning with abstention.

cs.LG↗

arXiv2021

Deep Learning with Label Differential Privacy

The Randomized Response (RR) algorithm is a classical technique to improve robustness in survey aggregation, and has been widely adopted in applications with differential privacy guarantees. We propose a novel algorithm, Randomized Response with Prior (RRWithPrior), which can provide more accurate results while maintaining the same level of privacy guaranteed by RR. We then apply RRWithPrior to learn neural networks with label differential privacy (LabelDP), and show that when only the label needs to be protected, the model performance can be significantly improved over the previous state-of-the-art private baselines. Moreover, we study different ways to obtain priors, which when used with RRWithPrior can additionally improve the model performance, further reducing the accuracy gap between private and non-private models. We complement the empirical results with theoretical analysis showing that LabelDP is provably easier than protecting both the inputs and labels.

cs.LG↗

arXiv2021

Fock-space perturbed relativistic coupled-cluster theory for electric dipole polarizability of one-valence atomic systems: Application to Al and In

We have developed a Fock-space relativistic coupled-cluster theory based method for the calculation of electric dipole polarizability of one-valence atoms and ions. We employ this method to compute the ground-state and spin-orbit coupled excited state electric dipole polarizability of Al and In. To check the quality of many-electron wavefunctions, we also compute the excitation energies of some low-lying states of Al and In. The effects of Breit interaction and QED corrections from the Uehling potential and the self-energy are included to improve the accuracy of $α$ further. Our recommended value of ground-state $α$ for both atoms are in good agreement with the previous theoretical results. From our computations, we find that more than 65\% of contributions come from the dipolar mixing of $3p$($5p$) with $3d$($5d$) and $4s$($6s$)-electrons for Al(In). The largest Breit and QED contributions are found to be 1.3\% and 0.6\%, respectively.

physics.atom-ph↗

arXiv2021

Differentially Private Aggregation in the Shuffle Model: Almost Central Accuracy in Almost a Single Message

The shuffle model of differential privacy has attracted attention in the literature due to it being a middle ground between the well-studied central and local models. In this work, we study the problem of summing (aggregating) real numbers or integers, a basic primitive in numerous machine learning tasks, in the shuffle model. We give a protocol achieving error arbitrarily close to that of the (Discrete) Laplace mechanism in the central model, while each user only sends $1 + o(1)$ short messages in expectation.

cs.CR↗

arXiv2021

Large-Scale Differentially Private BERT

In this work, we study the large-scale pretraining of BERT-Large with differentially private SGD (DP-SGD). We show that combined with a careful implementation, scaling up the batch size to millions (i.e., mega-batches) improves the utility of the DP-SGD step for BERT; we also enhance its efficiency by using an increasing batch size schedule. Our implementation builds on the recent work of [SVK20], who demonstrated that the overhead of a DP-SGD step is minimized with effective use of JAX [BFH+18, FJL18] primitives in conjunction with the XLA compiler [XLA17]. Our implementation achieves a masked language model accuracy of 60.5% at a batch size of 2M, for $ε= 5.36$. To put this number in perspective, non-private BERT models achieve an accuracy of $\sim$70%.

cs.LG↗

arXiv2021

Observation of ballistic upstream modes at fractional quantum Hall edges of graphene

The structure of edge modes at the boundary of quantum Hall (QH) phases forms the basis for understanding low energy transport properties. In particular, the presence of ``upstream'' modes, moving against the direction of charge current flow, is critical for the emergence of renormalized modes with exotic quantum statistics. Detection of excess noise at the edge is a smoking gun for the presence of upstream modes. Here we report on noise measurements at the edges of fractional QH (FQH) phases realized in dual graphite-gated bilayer graphene devices. A noiseless dc current is injected at one of the edge contacts, and the noise generated at contacts at $L= 4\,μ$m or $10\,μ$m away along the upstream direction is studied. For integer and particle-like FQH states, no detectable noise is measured. By contrast, for ``hole-conjugate'' FQH states, we detect a strong noise proportional to the injected current, unambiguously proving the existence of upstream modes. The noise magnitude remaining independent of length together with a remarkable agreement with our theoretical analysis demonstrates the ballistic nature of upstream energy transport, quite distinct from the diffusive propagation reported earlier in GaAs-based systems. Our investigation opens the door to the study of upstream transport in more complex geometries and in edges of non-Abelian phases in graphene.

cond-mat.mes-hall↗

arXiv2021

Google COVID-19 Vaccination Search Insights: Anonymization Process Description

This report describes the aggregation and anonymization process applied to the COVID-19 Vaccination Search Insights (published at http://goo.gle/covid19vaccinationinsights), a publicly available dataset showing aggregated and anonymized trends in Google searches related to COVID-19 vaccination. The applied anonymization techniques protect every user's daily search activity related to COVID-19 vaccinations with $(\varepsilon, δ)$-differential privacy for $\varepsilon = 2.19$ and $δ= 10^{-5}$.

cs.CR↗

arXiv2021

Private Counting from Anonymous Messages: Near-Optimal Accuracy with Vanishing Communication Overhead

Differential privacy (DP) is a formal notion for quantifying the privacy loss of algorithms. Algorithms in the central model of DP achieve high accuracy but make the strongest trust assumptions whereas those in the local DP model make the weakest trust assumptions but incur substantial accuracy loss. The shuffled DP model (Bittau et al., 2017; Erlingsson et al., 2019; Cheu et al., 2019) has recently emerged as a feasible middle ground between the central and local models, providing stronger trust assumptions than the former while promising higher accuracies than the latter. In this paper, we obtain practical communication-efficient algorithms in the shuffled DP model for two basic aggregation primitives used in machine learning: 1) binary summation, and 2) histograms over a moderate number of buckets. Our algorithms achieve accuracy that is arbitrarily close to that of central DP algorithms with an expected communication per user essentially matching what is needed without any privacy constraints! We demonstrate the practicality of our algorithms by experimentally comparing their performance to several widely-used protocols such as Randomized Response (Warner, 1965) and RAPPOR (Erlingsson et al., 2014).

cs.CR↗

arXiv2021

Shack Hartmann wavefront sensor with array of phase coded masks

Shack-Hartmann wavefront sensors (SHWS) are generally used to measure the wavefront curvature of light beams. Measurement accuracy and the sensitivity of these sensors are important factors for better wavefront sensing. In this study, we demonstrate a new type of SHWS with better measurement accuracy than the regular SHWS. The lenslet array in the regular SHWS is replaced with an array of coded phase masks and the principle of coded aperture correlation holography (COACH) is used for wavefront reconstruction. Sharper correlation peaks achieved by COACH improve the accuracy of the estimated local slopes of the measured wavefront and consequently improve the reconstruction accuracy of the overall wavefront. Experimental results confirm that the proposed method provides a lower mean square wavefront error by one order of magnitude in comparison to the regular SHWS.

physics.optics↗

arXiv2021

Locally Private k-Means in One Round

We provide an approximation algorithm for k-means clustering in the one-round (aka non-interactive) local model of differential privacy (DP). This algorithm achieves an approximation ratio arbitrarily close to the best non private approximation algorithm, improving upon previously known algorithms that only guarantee large (constant) approximation ratios. Furthermore, this is the first constant-factor approximation algorithm for k-means that requires only one round of communication in the local DP model, positively resolving an open question of Stemmer (SODA 2020). Our algorithmic framework is quite flexible; we demonstrate this by showing that it also yields a similar near-optimal approximation algorithm in the (one-round) shuffle DP model.

cs.DS↗

arXiv2021

X-ray Absorption Spectroscopy Study of La$_{1-y}$Sr$_{y}$Co$_{1-x}$Nb$_x$O$_3$

We use x-ray absorption spectroscopy to investigate the local structure and electronic properties of bulk La$_{1-y}$Sr$_{y}$Co$_{1-x}$Nb$_x$O$_3$ ($y=$ 2$x$ as LSCNO and $y=$ 0 as LCNO) samples. The x-ray absorption near-edge spectra (XANES) of LSCNO at Co K-edge affirm the valence state of Co in 3+. However, in the case of the LCNO, a subtle variation in the valence state of Co ions from 3+ to 2+ is evident with Nb substitution. The detailed analysis of the Fourier transform (FT) of the extended x-ray absorption fine structure (EXAFS) for the LSCNO samples exhibit the two groups of the bond-lengths owing to the Jahn-Teller (JT) distortion in the CoO$_6$ octahedra, which manifest that the Co$^{3+}$ ions exist in the intermediate spin-state (t$^5_{2g}$e$^1_g$) at room temperature. However, we find that the JT distortion is not present in LCNO samples for $x>$0.025 due to an increase in the high-spin Co$^{2+}$/Co$^{3+}$ ions accompanied by the Nb--induced structural transformation. Intriguingly, the La L$_3$-edge spectra for the LSCNO samples exhibit that the La ions exist in the trivalent state and the local disorder around La atoms decreases with Sr and Nb substitution. Interestingly, the simulated FT of the EXAFS spectra at the La L$_3$-edge demonstrates the three groups of La--O bond lengths, which exhibit a monotonous change with Sr and Nb substitution. Moreover, the XANES measured at Sr and Nb K-edges confirm their oxidation state to be in 2+ and 5+, respectively.

cond-mat.mtrl-sci↗

arXiv2021

Robust and Private Learning of Halfspaces

In this work, we study the trade-off between differential privacy and adversarial robustness under L2-perturbations in the context of learning halfspaces. We prove nearly tight bounds on the sample complexity of robust private learning of halfspaces for a large regime of parameters. A highlight of our results is that robust and private learning is harder than robust or private learning alone. We complement our theoretical analysis with experimental results on the MNIST and USPS datasets, for a learning algorithm that is both differentially private and adversarially robust.

cs.LG↗

arXiv2021

Vanishing thermal equilibration for hole-conjugate fractional quantum Hall states in graphene

Transport through edge-channels is responsible for conduction in quantum Hall (QH) phases. Topology dictates quantization of both charge and thermal transport coefficients. These turn out to approach robust quantized values when incoherent equilibration processes become dominant. Here, we report on measurements of both electrical and thermal conductances of integer and fractional quantum Hall (FQH) phases, realized in hBN encapsulated graphite gated bilayer graphene devices. Remarkably, for the complex edge at filling factors $ν=5/3$ and $ν=8/3$, which correspond to the paradigmatic hole-conjugate FQH phase $ν=2/3$ of the partially filled Landau level, we find vanishing thermal equilibration. This is striking, given that, at the same time, our results for the electrical conductance indicate efficient charge equilibration. These results are in accord with our theoretical analysis, pointing to a divergent thermal equilibration length in the limit of strong electrostatic interaction. Our results elucidate the subtle nature of the crossover from mesoscopic to robust topology-dominated transport in electronic two-dimensional topological phases.

cond-mat.mes-hall↗