arXiv subjects

Tara Javidi

Publications and source records attributed to Tara Javidi.

At least 73 records · Page 4Linked to original sources

arXiv2018

Efficient Video Understanding via Layered Multi Frame-Rate Analysis

One of the greatest challenges in the design of a real-time perception system for autonomous driving vehicles and drones is the conflicting requirement of safety (high prediction accuracy) and efficiency. Traditional approaches use a single frame rate for the entire system. Motivated by the observation that the lack of robustness against environmental factors is the major weakness of compact ConvNet architectures, we propose a dual frame-rate system that brings in the best of both worlds: A modulator stream that executes an expensive models robust to environmental factors at a low frame rate to extract slowly changing features describing the environment, and a prediction stream that executes a light-weight model at real-time to extract transient signals that describes particularities of the current frame. The advantage of our design is validated by our extensive empirical study, showing that our solution leads to consistent improvements using a variety of backbone architecture choice and input resolutions. These findings suggest multiple frame-rate systems as a promising direction in designing efficient perception for autonomous agents.

cs.CV↗

arXiv2018

DeepFense: Online Accelerated Defense Against Adversarial Deep Learning

Recent advances in adversarial Deep Learning (DL) have opened up a largely unexplored surface for malicious attacks jeopardizing the integrity of autonomous DL systems. With the wide-spread usage of DL in critical and time-sensitive applications, including unmanned vehicles, drones, and video surveillance systems, online detection of malicious inputs is of utmost importance. We propose DeepFense, the first end-to-end automated framework that simultaneously enables efficient and safe execution of DL models. DeepFense formalizes the goal of thwarting adversarial attacks as an optimization problem that minimizes the rarely observed regions in the latent feature space spanned by a DL network. To solve the aforementioned minimization problem, a set of complementary but disjoint modular redundancies are trained to validate the legitimacy of the input samples in parallel with the victim DL model. DeepFense leverages hardware/software/algorithm co-design and customized acceleration to achieve just-in-time performance in resource-constrained settings. The proposed countermeasure is unsupervised, meaning that no adversarial sample is leveraged to train modular redundancies. We further provide an accompanying API to reduce the non-recurring engineering cost and ensure automated adaptation to various platforms. Extensive evaluations on FPGAs and GPUs demonstrate up to two orders of magnitude performance improvement while enabling online adversarial sample detection.

cs.CR↗

arXiv2018

Active Learning with Logged Data

We consider active learning with logged data, where labeled examples are drawn conditioned on a predetermined logging policy, and the goal is to learn a classifier on the entire population, not just conditioned on the logging policy. Prior work addresses this problem either when only logged data is available, or purely in a controlled random experimentation setting where the logged data is ignored. In this work, we combine both approaches to provide an algorithm that uses logged data to bootstrap and inform experimentation, thus achieving the best of both worlds. Our work is inspired by a connection between controlled random experimentation and active learning, and modifies existing disagreement-based active learning algorithms to exploit logged data.

cs.LG↗

arXiv2018

Dynamic Cloud Network Control under Reconfiguration Delay and Cost

Network virtualization and programmability allow operators to deploy a wide range of services over a common physical infrastructure and elastically allocate cloud and network resources according to changing requirements. While the elastic reconfiguration of virtual resources enables dynamically scaling capacity in order to support service demands with minimal operational cost, reconfiguration operations make resources unavailable during a given time period and may incur additional cost. In this paper, we address the dynamic cloud network control problem under non-negligible reconfiguration delay and cost. We show that while the capacity region remains unchanged regardless of the reconfiguration delay/cost values, a reconfiguration-agnostic policy may fail to guarantee throughput-optimality and minimum cost under nonzero reconfiguration delay/cost. We then present an adaptive dynamic cloud network control policy that allows network nodes to make local flow scheduling and resource allocation decisions while controlling the frequency of reconfiguration in order to support any input rate in the capacity region and achieve arbitrarily close to minimum cost for any finite reconfiguration delay/cost values.

cs.NI↗

arXiv2018

Gaussian Process bandits with adaptive discretization

In this paper, the problem of maximizing a black-box function $f:\mathcal{X} \to \mathbb{R}$ is studied in the Bayesian framework with a Gaussian Process (GP) prior. In particular, a new algorithm for this problem is proposed, and high probability bounds on its simple and cumulative regret are established. The query point selection rule in most existing methods involves an exhaustive search over an increasingly fine sequence of uniform discretizations of $\mathcal{X}$. The proposed algorithm, in contrast, adaptively refines $\mathcal{X}$ which leads to a lower computational complexity, particularly when $\mathcal{X}$ is a subset of a high dimensional Euclidean space. In addition to the computational gains, sufficient conditions are identified under which the regret bounds of the new algorithm improve upon the known results. Finally an extension of the algorithm to the case of contextual bandits is proposed, and high probability bounds on the contextual regret are presented.

stat.ML↗

arXiv2017

Improved Target Acquisition Rates with Feedback Codes

This paper considers the problem of acquiring an unknown target location (among a finite number of locations) via a sequence of measurements, where each measurement consists of simultaneously probing a group of locations. The resulting observation consists of a sum of an indicator of the target's presence in the probed region, and a zero mean Gaussian noise term whose variance is a function of the measurement vector. An equivalence between the target acquisition problem and channel coding over a binary input additive white Gaussian noise (BAWGN) channel with state and feedback is established. Utilizing this information theoretic perspective, a two-stage adaptive target search strategy based on the sorted Posterior Matching channel coding strategy is proposed. Furthermore, using information theoretic converses, the fundamental limits on the target acquisition rate for adaptive and non-adaptive strategies are characterized. As a corollary to the non-asymptotic upper bound of the expected number of measurements under the proposed two-stage strategy, and to non-asymptotic lower bound of the expected number of measurements for optimal non-adaptive search strategy, a lower bound on the adaptivity gain is obtained. The adaptivity gain is further investigated in different asymptotic regimes of interest.

cs.IT↗

arXiv2017

Percentile Policies for Tracking of Markovian Random Processes with Asymmetric Cost and Observation

Motivated by wide-ranging applications such as video delivery over networks using Multiple Description Codes, congestion control, and inventory management, we study the state-tracking of a Markovian random process with a known transition matrix and a finite ordered state set. The decision-maker must select a state as an action at each time step to minimize the total expected cost. The decision-maker is faced with asymmetries both in cost and observation: in case the selected state is less than the actual state of the Markovian process, an under-utilization cost occurs and only partial observation about the actual state is revealed; otherwise, the decision incurs an over-utilization cost and reveals full information about the actual state. We can formulate this problem as a Partially Observable Markov Decision Process which can be expressed as a dynamic program based on the last full observed state and the time of full observation. This formulation determines the sequence of actions to be taken between any two consecutive full observations of the actual state. However, this DP grows exponentially in the number of states, with little hope for a computationally feasible solution. We present an interesting class of computationally tractable policies with a percentile structure. A generalization of binary search, this class of policies attempt at any given time to reduce the uncertainty by a given percentage. Among all percentile policies, we search for the one with the minimum expected cost. The result of this search is a heuristic policy which we evaluate through numerical simulations. We show that it outperforms the myopic policies and under some conditions performs close to the optimal policies. Furthermore, we derive a lower bound on the cost of the optimal policy which can be computed with low complexity and give a measure for how close our heuristic policy is to the optimal policy.

cs.IT↗

arXiv2017

Heavy Traffic Queue Length Behavior in Switches with Reconfiguration Delay

Optical switches have been drawing attention due to their large data bandwidth and low power consumption. However, scheduling policies need to account for the schedule reconfiguration delay of optical switches to achieve good performance. The Adaptive MaxWeight policy achieves optimal throughput for switches with nonzero reconfiguration delay, and has been shown in simulation to have good delay performance. In this paper, we analyze the queue length behavior of a switch with nonzero reconfiguration delay operating under the Adaptive MaxWeight. We first show that the Adaptive MaxWeight policy exhibits a weak state space collapse behavior in steady-state, which could be viewed as an inheritance of the MaxWeight policy in a switch with zero reconfiguration delay. We then use the weak state space collapse result to obtain a steady state delay bound under the Adaptive MaxWeight algorithm in heavy traffic by applying a recently developed drift technique. The resulting delay bound is dependent on the expected schedule duration. We then derive the relation between the expected schedule duration and the steady state queue length through drift analysis, and obtain asymptotically tight queue length bounds in the heavy traffic regime.

math.OC↗

arXiv2016

Searching with Measurement Dependent Noise

Consider a target moving at a constant velocity on a unit-circumference circle, starting at an arbitrary location. To acquire the target, any region of the circle can be probed to obtain a noisy measurement of the target's presence, where the noise level increases with the size of the probed region. We are interested in the expected time required to find the target to within some given resolution and error probability. For a known velocity, we characterize the optimal tradeoff between time and resolution, and show that in contrast to the well studied case of constant measurement noise, measurement dependent noise incurs a multiplicative gap in the targeting rate between adaptive and non-adaptive search strategies. Moreover, our adaptive strategy attains the optimal rate-reliability tradeoff. We further show that for optimal non-adaptive search, accounting for an unknown velocity incurs a factor of at least two in the targeting rate.

cs.IT↗

arXiv2016

Fully-adaptive Feature Sharing in Multi-Task Networks with Applications in Person Attribute Classification

Multi-task learning aims to improve generalization performance of multiple prediction tasks by appropriately sharing relevant information across them. In the context of deep neural networks, this idea is often realized by hand-designed network architectures with layers that are shared across tasks and branches that encode task-specific features. However, the space of possible multi-task deep architectures is combinatorially large and often the final architecture is arrived at by manual exploration of this space subject to designer's bias, which can be both error-prone and tedious. In this work, we propose a principled approach for designing compact multi-task deep learning architectures. Our approach starts with a thin network and dynamically widens it in a greedy manner during training using a novel criterion that promotes grouping of similar tasks together. Our Extensive evaluation on person attributes classification tasks involving facial and clothing attributes suggests that the models produced by the proposed method are fast, compact and can closely match or exceed the state-of-the-art accuracy from strong baselines by much more expensive models.

cs.CV↗

arXiv2016

Active Learning from Imperfect Labelers

We study active learning where the labeler can not only return incorrect labels but also abstain from labeling. We consider different noise and abstention conditions of the labeler. We propose an algorithm which utilizes abstention responses, and analyze its statistical consistency and query complexity under fairly natural assumptions on the noise and abstention rate of the labeler. This algorithm is adaptive in a sense that it can automatically request less queries with a more informed or less noisy labeler. We couple our algorithm with lower bounds to show that under some technical conditions, it achieves nearly optimal query complexity.

cs.LG↗

arXiv2016

Optimal Pricing to Manage Electric Vehicles in Coupled Power and Transportation Networks

We study the system-level effects of the introduction of large populations of Electric Vehicles on the power and transportation networks. We assume that each EV owner solves a decision problem to pick a cost-minimizing charge and travel plan. This individual decision takes into account traffic congestion in the transportation network, affecting travel times, as well as as congestion in the power grid, resulting in spatial variations in electricity prices for battery charging. We show that this decision problem is equivalent to finding the shortest path on an "extended" transportation graph, with virtual arcs that represent charging options. Using this extended graph, we study the collective effects of a large number of EV owners individually solving this path planning problem. We propose a scheme in which independent power and transportation system operators can collaborate to manage each network towards a socially optimum operating point while keeping the operational data of each system private. We further study the optimal reserve capacity requirements for pricing in the absence of such collaboration. We showcase numerically that a lack of attention to interdependencies between the two infrastructures can have adverse operational effects.

eess.SY↗

arXiv2016

Mobile Relays for Smart Cities: Mathematical Proofs

The increasing number of connected vehicles in densely populated urban areas provides an interesting opportunity to counteract the high wireless data demands in high density and highly mobile scenarios. The idea is to support the macro base station (BS) with a secondary communication tier composed of a set of smart and connected vehicles that are in movement in the urban area. As a first step towards a comprehensive cost-benefit analysis of this architecture, this paper considers the case where these vehicles are equipped with femto-mobile Access Points (fmAPs) and constitute a mobile out-of-band relay infrastructure. In particular, three techniques to select an fmAP (if more than one is available) are proposed and the maximal feasible gain in the packet delivery rate and data rate as a function of the vehicle density, average vehicle speeds, handoff overhead cost, as well as physical layer parameters is characterized. The analytical and simulation results provide a first benchmark characterizing this architecture and the definition of guidelines for its future realistic study and implementation.

cs.IT↗

arXiv2016

Synchronization in Networked Systems with Parameter Mismatch: Adaptive Decentralized and Distributed Controls

Here, we study the ultimately bounded stability of network of mismatched systems using Lyapunov direct method. We derive an upper bound on the norm of the error of network states from its average states, which it achieves in finite time. Then, we devise a decentralized compensator to asymptotically pin the network of mismatched systems to a desired trajectory. Next, we design distributed estimators to compensate for the mismatched parameters performances of adaptive decentralized and distributed compensations are analyzed. Our analytical results are verified by several simulations in a network of globally connected Lorenz oscillators.

eess.SY↗

arXiv2016

Reliability of Sequential Hypothesis Testing Can Be Achieved by an Almost-Fixed-Length Test

The maximum type-I and type-II error exponents associated with the newly introduced almost-fixed-length hypothesis testing is characterized. In this class of tests, the decision-maker declares the true hypothesis almost always after collecting a fixed number of samples $n$; however in very rare cases with exponentially small probability the decision maker is allowed to collect another set of samples (no more than polynomial in $n$). This class of hypothesis tests are shown to bridge the gap between the classical hypothesis testing with a fixed sample size and the sequential hypothesis testing, and improve the trade-off between type-I and type-II error exponents.

cs.IT↗

arXiv2016

Social Learning and Distributed Hypothesis Testing

This paper considers a problem of distributed hypothesis testing and social learning. Individual nodes in a network receive noisy local (private) observations whose distribution is parameterized by a discrete parameter (hypotheses). The conditional distributions are known locally at the nodes, but the true parameter/hypothesis is not known. An update rule is analyzed in which nodes first perform a Bayesian update of their belief (distribution estimate) of the parameter based on their local observation, communicate these updates to their neighbors, and then perform a "non-Bayesian" linear consensus using the log-beliefs of their neighbors. In this paper we show that under mild assumptions, the belief of any node in any incorrect hypothesis converges to zero exponentially fast, and we characterize the exponential rate of learning which is given in terms of the network structure and the divergences between the observations' distributions. Our main result is the concentration property established on the rate of convergence.

math.ST↗

arXiv2016

Adaptive Object Detection Using Adjacency and Zoom Prediction

State-of-the-art object detection systems rely on an accurate set of region proposals. Several recent methods use a neural network architecture to hypothesize promising object locations. While these approaches are computationally efficient, they rely on fixed image regions as anchors for predictions. In this paper we propose to use a search strategy that adaptively directs computational resources to sub-regions likely to contain objects. Compared to methods based on fixed anchor locations, our approach naturally adapts to cases where object instances are sparse and small. Our approach is comparable in terms of accuracy to the state-of-the-art Faster R-CNN approach while using two orders of magnitude fewer anchors on average. Code is publicly available.

cs.CV↗

arXiv2015

Bounded Stability in Networked Systems with Parameter Mismatch and Adaptive Decentralized Estimation

Here, we study the ultimately bounded stability of network of mismatched systems using Lyapunov direct method. The upper bound on the error of oscillators from the center of the neighborhood is derived. Then the performance of an adaptive compensation via decentralized control is analyzed. Finally, the analytical results for a network of globally connected Lorenz oscillators are verified.

eess.SY↗