arXiv subjects

Upamanyu Madhow

Publications and source records attributed to Upamanyu Madhow.

At least 37 records · Page 2Linked to original sources

arXiv2020

A design framework for all-digital mmWave massive MIMO with per-antenna nonlinearities

Millimeter wave MIMO combines the benefits of compact antenna arrays with a large number of elements and massive bandwidths, so that fully digital beamforming has the potential of supporting a large number of simultaneous users with {\it per user} data rates of multiple gigabits/sec (Gbps). In this paper, we develop an analytical model for the impact of nonlinearities in such a system, and illustrate its utility in providing hardware design guidelines regarding two key challenges: the low available precision of analog-to-digital conversion at high sampling rates, and nonlinearities in ultra-high speed radio frequency (RF) and baseband circuits. We consider linear minimum mean square error (LMMSE) reception for a multiuser MIMO uplink, and provide performance guarantees based on two key concepts: (a) summarization of the impact of per-antenna nonlinearities via a quantity that we term the "intrinsic SNR", (b) using linear MMSE performance in an ideal system without nonlinearities to bound that in our non-ideal system. For our numerical results, we employ nominal parameters corresponding to outdoor picocells operating at a carrier frequency of 140 GHz, with a data rate of 10 Gbps per user.

eess.SP↗

arXiv2020

Polarizing Front Ends for Robust CNNs

The vulnerability of deep neural networks to small, adversarially designed perturbations can be attributed to their "excessive linearity." In this paper, we propose a bottom-up strategy for attenuating adversarial perturbations using a nonlinear front end which polarizes and quantizes the data. We observe that ideal polarization can be utilized to completely eliminate perturbations, develop algorithms to learn approximately polarizing bases for data, and investigate the effectiveness of the proposed strategy on the MNIST and Fashion MNIST datasets.

stat.ML↗

arXiv2019

Phase noise in modular millimeter wave massive MIMO

This paper investigates the effect of oscillator phase noise on a multiuser millimeter wave (mmWave) massive MIMO uplink as we scale up the number of base station antennas, fixing the load factor, defined as the ratio of the number of simultaneous users to the number of base station antennas. We consider a modular approach in which the base station employs an array of subarrays, or "tiles." Each tile supports a fixed number of antennas, and can therefore be implemented using a separate radio frequency integrated circuit (RFIC), with synchronization across tiles accomplished by employing a phased locked loop in each tile to synthesize an on-chip oscillator at the carrier frequency by locking on to a common lower frequency reference clock. Assuming linear minimum mean squared error (LMMSE) multiuser detection, we provide an analytical framework that can be used to specify the required power spectral density (PSD) mask for phase noise for a target system performance. Our analysis for the phase noise at the output of the LMMSE receiver indicates two distinct effects: self-noise for each user which is inversely proportional to the number of tiles, and cross-talk between users which is insensitive to the number of tiles, and is proportional to the load factor. These analytical predictions, verified by simulations for a 140 GHz system targeting a per-user data rate of 10 Gbps, show that tiling is a robust approach for scaling. Numerical results for our proposed design approach yield relatively relaxed specifications for phase noise PSD masks.

eess.SP↗

arXiv2019

Robust Wireless Fingerprinting via Complex-Valued Neural Networks

A "wireless fingerprint" which exploits hardware imperfections unique to each device is a potentially powerful tool for wireless security. Such a fingerprint should be able to distinguish between devices sending the same message, and should be robust against standard spoofing techniques. Since the information in wireless signals resides in complex baseband, in this paper, we explore the use of neural networks with complex-valued weights to learn fingerprints using supervised learning. We demonstrate that, while there are potential benefits to using sections of the signal beyond just the preamble to learn fingerprints, the network cheats when it can, using information such as transmitter ID (which can be easily spoofed) to artificially inflate performance. We also show that noise augmentation by inserting additional white Gaussian noise can lead to significant performance gains, which indicates that this counter-intuitive strategy helps in learning more robust fingerprints. We provide results for two different wireless protocols, WiFi and ADS-B, demonstrating the effectiveness of the proposed method.

eess.SP↗

arXiv2019

Design of Large Effective Apertures for Millimeter Wave Systems using a Sparse Array of Subarrays

We investigate synthesis of a large effective aperture using a sparse array of subarrays. We employ a multi-objective optimization framework for placement of subarrays within a prescribed area dictated by form factor constraints, trading off the smaller beam width obtained by spacing out the subarrays against the grating and side lobes created by sparse placement. We assess the performance of our designs for the fundamental problem of bearing estimation for one or more sources, comparing performance against estimation-theoretic bounds. Our tiled architecture is motivated by recent progress in low-cost hardware realizations of moderately sized antenna arrays (which play the role of subarrays) in the millimeter wave band, and our numerical examples are based on 16-element (4x4) subarrays in the 60 GHz unlicensed band.

eess.SP↗

arXiv2019

Interference management and capacity analysis for mm-wave picocells in urban canyons

Millimeter (mm) wave picocellular networks are a promising approach for delivering the 1000-fold capacity increase required to keep up with projected demand for wireless data: the available bandwidth is orders of magnitude larger than that in existing cellular systems, and the small carrier wavelength enables the realization of highly directive antenna arrays in compact form factor, thus drastically increasing spatial reuse. In this paper, we carry out an interference analysis for mm-wave picocells in an urban canyon with a dense deployment of base stations. Each base station sector can serve multiple simultaneous users, which implies that both intra- and inter-cell interference must be managed. We propose a \textit{cross-layer} approach to interference management based on (i) suppressing interference at the physical layer and (ii) managing the residual interference at the medium access control layer. We provide an estimate of network capacity and establish that 1000-fold increase relative to conventional LTE cellular networks is indeed feasible.

eess.SP↗

arXiv2018

Noncoherent compressive channel estimation for mm-wave massive MIMO

Millimeter (mm) wave massive MIMO has the potential for delivering orders of magnitude increases in mobile data rates, with compact antenna arrays providing narrow steerable beams for unprecedented levels of spatial reuse. A fundamental technical bottleneck, however, is rapid spatial channel estimation and beam adaptation in the face of mobility and blockage. Recently proposed compressive techniques which exploit the sparsity of mm wave channels are a promising approach to this problem, with overhead scaling linearly with the number of dominant paths and logarithmically with the number of array elements. Further, they can be implemented with RF beamforming with low-precision phase control. However, these methods make implicit assumptions on long-term phase coherence that are not satisfied by existing hardware. In this paper, we propose and evaluate a noncoherent compressive channel estimation technique which can estimate a sparse spatial channel based on received signal strength (RSS) alone, and is compatible with off-the-shelf hardware. The approach is based on cascading phase retrieval (i.e., recovery of complex-valued measurements from RSS measurements, up to a scalar multiple) with coherent compressive estimation. While a conventional cascade scheme would multiply two measurement matrices to obtain an overall matrix whose entries are in a continuum, a key novelty in our scheme is that we constrain the overall measurement matrix to be implementable using coarsely quantized pseudorandom phases, employing a virtual decomposition of the matrix into a product of measurement matrices for phase retrieval and compressive estimation. Theoretical and simulation results show that our noncoherent method scales almost as well with array size as its coherent counterpart, thus inheriting the scalability and low overhead of the latter.

eess.SP↗

arXiv2018

Joint Routing and Resource Allocation for Millimeter Wave Picocellular Backhaul

Picocellular architectures are essential for providing the spatial reuse required to satisfy the ever-increasing demand for mobile data. A key deployment challenge is to provide backhaul connections with sufficiently high data rate. Providing wired support (e.g., using optical fiber) to pico base stations deployed opportunistically on lampposts and rooftops is impractical, hence wireless backhaul becomes an attractive approach. A multihop mesh network comprised of directional millimeter wave links is considered here for this purpose. The backhaul design problem is formulated as one of joint routing and resource allocation, accounting for mutual interference across simultaneously active links. A computationally tractable formulation is developed by leveraging the localized nature of interference and the provable existence of a sparse optimal allocation. Numerical results are provided for millimeter (mm) wave mesh networks, which are well suited for scaling backhaul data rates due to abundance of spectrum, and the ability to form highly directional, electronically steerable beams.

eess.SP↗

arXiv2018

Geometry-constrained Degrees of Freedom Analysis for Imaging Systems: Monostatic and Multistatic

In this paper, we develop a theoretical framework for analyzing the measurable information content of an unknown scene through an active electromagnetic imaging array. We consider monostatic and multistatic array architectures in a one-dimensional setting. Our main results include the following: (a) we introduce the space-bandwidth product (SBP), and show that, under the Born approximation, it provides an accurate prediction of the number of the degrees of freedom (DoF) as constrained by the geometry of the scene and the imaging system; (b) we show that both monostatic and multistatic architectures have the same number of DoF; (c) we show that prior DoF analysis based on the more restrictive Fresnel approximation are obtained by specializing our results; (d) we investigate matched-filter (back-propagation) and pseudoinverse image reconstruction schemes, and analyze the achievable resolution through these methods. Our analytical framework opens up new avenues to investigate image formation techniques that aim to reconstruct the reflectivity function of the scene by solving an inverse scattering problem, and provides insights on achievable resolution. For example, we show that matched-filter reconstruction leads to a significant resolution loss for multistatic architectures.

cs.IT↗

arXiv2018

Combating Adversarial Attacks Using Sparse Representations

It is by now well-known that small adversarial perturbations can induce classification errors in deep neural networks (DNNs). In this paper, we make the case that sparse representations of the input data are a crucial tool for combating such attacks. For linear classifiers, we show that a sparsifying front end is provably effective against $\ell_{\infty}$-bounded attacks, reducing output distortion due to the attack by a factor of roughly $K / N$ where $N$ is the data dimension and $K$ is the sparsity level. We then extend this concept to DNNs, showing that a "locally linear" model can be used to develop a theoretical foundation for crafting attacks and defenses. Experimental results for the MNIST dataset show the efficacy of the proposed sparsifying front end.

stat.ML↗

arXiv2018

Sparsity-based Defense against Adversarial Attacks on Linear Classifiers

Deep neural networks represent the state of the art in machine learning in a growing number of fields, including vision, speech and natural language processing. However, recent work raises important questions about the robustness of such architectures, by showing that it is possible to induce classification errors through tiny, almost imperceptible, perturbations. Vulnerability to such "adversarial attacks", or "adversarial examples", has been conjectured to be due to the excessive linearity of deep networks. In this paper, we study this phenomenon in the setting of a linear classifier, and show that it is possible to exploit sparsity in natural data to combat $\ell_{\infty}$-bounded adversarial perturbations. Specifically, we demonstrate the efficacy of a sparsifying front end via an ensemble averaged analysis, and experimental results for the MNIST handwritten digit database. To the best of our knowledge, this is the first work to show that sparsity provides a theoretically rigorous framework for defense against adversarial attacks.

stat.ML↗

arXiv2018

On the information in spike timing: neural codes derived from polychronous groups

There is growing evidence regarding the importance of spike timing in neural information processing, with even a small number of spikes carrying information, but computational models lag significantly behind those for rate coding. Experimental evidence on neuronal behavior is consistent with the dynamical and state dependent behavior provided by recurrent connections. This motivates the minimalistic abstraction investigated in this paper, aimed at providing insight into information encoding in spike timing via recurrent connections. We employ information-theoretic techniques for a simple reservoir model which encodes input spatiotemporal patterns into a sparse neural code, translating the polychronous groups introduced by Izhikevich into codewords on which we can perform standard vector operations. We show that the distance properties of the code are similar to those for (optimal) random codes. In particular, the code meets benchmarks associated with both linear classification and capacity, with the latter scaling exponentially with reservoir size.

q-bio.NC↗

arXiv2018

RF Source Seeking using Frequency Measurements

In this paper, we consider a problem motivated by search-and-rescue applications, where an unmanned aerial vehicle (UAV) seeks to approach the vicinity of a distant quasi-stationary radio frequency (RF) emitter surrounded by local scatterers. The UAV employs only measurements of the Doppler frequency of the received RF signal, along with its own bearing, to continuously adapt its trajectory. We propose and evaluate a trajectory planning approach that addresses technical difficulties such as the unknown carrier frequency offset between the emitter and the UAV's receiver, the frequency drifts of the local oscillators over time, the direction ambiguity in Doppler, and the noise in the observations. For the initial trajectory, the UAV estimates the direction of the emitter using a circular motion, which resolves direction ambiguity. The trajectory is then continuously adapted using feedback from frequency measurements obtained by perturbing the bearing around the current trajectory. We show that the proposed algorithm converges to the vicinity of the emitter, and illustrate its efficacy using simulations.

eess.SP↗

arXiv2018

Distributed Base Station: A Concept System for Long-Range Broadband Wireless Access

We propose a concept system termed distributed base station (DBS), which enables distributed transmit beamforming at large carrier wavelengths to achieve significant range extension and/or increased downlink data rate, providing a low-cost infrastructure for applications such as rural broadband. We consider a frequency division duplexed (FDD) system, using feedback from the receiver to achieve the required phase coherence. At a given range, $N$ cooperating transmitters can achieve $N^2$-fold increase in received power compared to that for a single transmitters, and feedback-based algorithms with near-ideal performance have been prototyped. In this paper, however, we identify and address key technical issues in translating such power gains into range extension via a DBS. First, to combat the drop in per-node SNR with extended range, we design a feedback-based adaptation strategy that is suitably robust to noise. Second, to utilize available system bandwidth, we extend narrowband adaptation algorithms to wideband channels through interpolation over OFDM subcarriers. Third, we observe that the feedback channel may become a bottleneck unless sophisticated distributed reception strategies are employed, but show that acceptable performance can still be obtained with standard uplink reception if channel time variations are slow enough. We quantify system performance compactly via outage capacity analyses.

cs.IT↗

arXiv2016

Learning Sparse, Distributed Representations using the Hebbian Principle

The "fire together, wire together" Hebbian model is a central principle for learning in neuroscience, but surprisingly, it has found limited applicability in modern machine learning. In this paper, we take a first step towards bridging this gap, by developing flavors of competitive Hebbian learning which produce sparse, distributed neural codes using online adaptation with minimal tuning. We propose an unsupervised algorithm, termed Adaptive Hebbian Learning (AHL). We illustrate the distributed nature of the learned representations via output entropy computations for synthetic data, and demonstrate superior performance, compared to standard alternatives such as autoencoders, in training a deep convolutional net on standard image datasets.

cs.LG↗

arXiv2016

Newtonized Orthogonal Matching Pursuit: Frequency Estimation over the Continuum

We propose a fast sequential algorithm for the fundamental problem of estimating frequencies and amplitudes of a noisy mixture of sinusoids. The algorithm is a natural generalization of Orthogonal Matching Pursuit (OMP) to the continuum using Newton refinements, and hence is termed Newtonized OMP (NOMP). Each iteration consists of two phases: detection of a new sinusoid, and sequential Newton refinements of the parameters of already detected sinusoids. The refinements play a critical role in two ways: (1) sidestepping the potential basis mismatch from discretizing a continuous parameter space, (2) providing feedback for locally refining parameters estimated in previous iterations. We characterize convergence, and provide a Constant False Alarm Rate (CFAR) based termination criterion. By benchmarking against the Cramer Rao Bound, we show that NOMP achieves near-optimal performance under a variety of conditions. We compare the performance of NOMP with classical algorithms such as MUSIC and more recent Atomic norm Soft Thresholding (AST) and Lasso algorithms, both in terms of frequency estimation accuracy and run time.

cs.IT↗

arXiv2015

Compressive spectral embedding: sidestepping the SVD

Spectral embedding based on the Singular Value Decomposition (SVD) is a widely used "preprocessing" step in many learning tasks, typically leading to dimensionality reduction by projecting onto a number of dominant singular vectors and rescaling the coordinate axes (by a predefined function of the singular value). However, the number of such vectors required to capture problem structure grows with problem size, and even partial SVD computation becomes a bottleneck. In this paper, we propose a low-complexity it compressive spectral embedding algorithm, which employs random projections and finite order polynomial expansions to compute approximations to SVD-based embedding. For an m times n matrix with T non-zeros, its time complexity is O((T+m+n)log(m+n)), and the embedding dimension is O(log(m+n)), both of which are independent of the number of singular vectors whose effect we wish to capture. To the best of our knowledge, this is the first work to circumvent this dependence on the number of singular vectors for general SVD-based embeddings. The key to sidestepping the SVD is the observation that, for downstream inference tasks such as clustering and classification, we are only interested in using the resulting embedding to evaluate pairwise similarity metrics derived from the euclidean norm, rather than capturing the effect of the underlying matrix on arbitrary vectors as a partial SVD tries to do. Our numerical results on network datasets demonstrate the efficacy of the proposed method, and motivate further exploration of its application to large-scale inference tasks.

stat.ML↗

arXiv2015

Compressive channel estimation and tracking for large arrays in mm wave picocells

We propose and investigate a compressive architecture for estimation and tracking of sparse spatial channels in millimeter (mm) wave picocellular networks. The base stations are equipped with antenna arrays with a large number of elements (which can fit within compact form factors because of the small carrier wavelength) and employ radio frequency (RF) beamforming, so that standard least squares adaptation techniques (which require access to individual antenna elements) are not applicable. We focus on the downlink, and show that "compressive beacons," transmitted using pseudorandom phase settings at the base station array, and compressively processed using pseudorandom phase settings at the mobile array, provide information sufficient for accurate estimation of the two-dimensional (2D) spatial frequencies associated with the directions of departure of the dominant rays from the base station, and the associated complex gains. This compressive approach is compatible with coarse phase-only control, and is based on a near-optimal sequential algorithm for frequency estimation which can exploit the geometric continuity of the channel across successive beaconing intervals to reduce the overhead to less than 1% even for very large (32 x 32) arrays. Compressive beaconing is essentially omnidirectional, and hence does not enjoy the SNR and spatial reuse benefits of beamforming obtained during data transmission. We therefore discuss system level design considerations for ensuring that the beacon SNR is sufficient for accurate channel estimation, and that inter-cell beacon interference is controlled by an appropriate reuse scheme.

cs.IT↗