Alvaro Velasquez - Academia.edu (original) (raw)

Papers by Alvaro Velasquez

Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulm... more Convolutional neural networks (CNN) are now being widely used for classifiying and detecting pulmonary abnormalities in chest radiographs. Two complementary generalization properties of CNNs, translation invariance and equivariance, are particularly useful in detecting manifested abnormalities associated with pulmonary disease, regardless of their spatial locations within the image. However, these properties also come with the loss of exact spatial information and global relative positions of abnormalities detected in local regions. Global relative positions of such abnormalities may help distinguish similar conditions, such as COVID-19 and viral pneumonia. In such instances, a global attention mechanism is needed, which CNNs do not support in their traditional architectures that aim for generalization afforded by translation invariance and equivariance. Vision Transformers provide a global attention mechanism, but lack translation invariance and equivariance, requiring significantl...

This paper considers a resilient high-dimensional constrained consensus problem and studies a res... more This paper considers a resilient high-dimensional constrained consensus problem and studies a resilient distributed algorithm for complete graphs. For convex constrained sets with a singleton intersection, a sufficient condition on feasibility redundancy and set regularity for reaching a desired consensus exponentially fast in the presence of Byzantine agents is derived, which can be directly applied to polyhedral sets. A necessary condition on feasibility redundancy for the resilient constrained consensus problem to be solvable is also provided.

Proposed for presentation at the Conference on Magnetism and Magnetic Materials held November 2-6, 2020 in Virtual., 2020

International Joint Conference on Artificial Intelligence, 2020

2021 60th IEEE Conference on Decision and Control (CDC), 2021

In this paper, we consider the cooperative control problem for a class of discrete-time nonlinear... more In this paper, we consider the cooperative control problem for a class of discrete-time nonlinear multiagent systems with the objective of minimizing a group cost functional. A multiagent Hamilton-Jacobi-Bellman (HJB) equation is first derived and then a new consensus-based value iteration algorithm is proposed to seek the online approximate solution to multiagent HJB. Neural networks are employed to parameterize the state value functions for individual agents, and a novel adaptive law for updating neural network weights is proposed based on the estimation of several global terms. The proposed local information based cooperative control is based on the minimization of the overall cost functional which is the sum of all individual agents’ cost functionals. Numerical simulations show the effectiveness of the proposed design.

2018 IEEE 36th International Conference on Computer Design (ICCD), 2018

A novel in-memory computing design that can compute single-source reachability and transitive clo... more A novel in-memory computing design that can compute single-source reachability and transitive closure of graphs is introduced. The proposed design leverages the parallel flow of information in three-dimensional crosspoint memories and can be implemented using memories with two layers of 1-diode 1-resistor (1D1R) interconnects. Our logic-in-memory designs mitigate the infamous memory-processor bottleneck characteristic of John von Neumann architectures and have runtime complexities of O(n) and O(n^2) using O(n^2) memory cells for the single-source reachability and transitive closure problems, respectively, where n is the number of nodes in the graph. This work builds upon preliminary results presented in [1].

2017 IEEE International Symposium on Circuits and Systems (ISCAS), 2017

Energy concerns, the infamous memory wall, and the enormous data deluge of the current big-data a... more Energy concerns, the infamous memory wall, and the enormous data deluge of the current big-data age have made the integration of processing and memory elements into a very appealing paradigm. In this paper, we focus on a computation-in-memory solution to the problem of multiplying a set of Boolean matrices, also known as Boolean matrix chain multiplication (BMCM). This is a fundamental computational task with applications in graph theory, group testing, data compression, and digital signal processing. In particular, we propose a framework for mapping arbitrary instances of BMCM to a 3-dimensional (3D) crossbar memory architecture consisting of 1-diode 1-resistor (1D1R) structures.

International Joint Conference on Artificial Intelligence, 2019

2018 Design, Automation & Test in Europe Conference & Exhibition (DATE), 2018

Lecture Notes in Computer Science, 2018

Observationofchemicalstateofsolid-solutioncarboninalow-carbonsteelwastriedbyC-Knear-edgex-rayabso... more Observationofchemicalstateofsolid-solutioncarboninalow-carbonsteelwastriedbyC-Knear-edgex-rayabsorptionfinestructurespectra measurement. In addition, the wavelength dependence of the photoelectron spectrum on the surface of the bulk steel was evaluated, and the contamination and oxidation layer of 3 nm in thickness on the surface of the steel was found. As a result, it was possible to observe the chemical state change of carbon existing in the bulk iron located deeper than the oxidation and contamination layer, by evaluating the difference spectra between the sample and the reference. Furthermore, by evaluating the shape change of the difference spectra depending on the heat treatment time, this study suggested that the chemical state of carbon in bulk iron changes with heat treatment.

Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, 2019

In this paper, we introduce the Steady-State Policy Synthesis (SSPS) problem which consists of ﬁn... more In this paper, we introduce the Steady-State Policy Synthesis (SSPS) problem which consists of ﬁnding a stochastic decision-making policy that maximizes expected rewards while satisfying a set of asymptotic behavioral speciﬁcations. These speciﬁcations are determined by the steady-state probability distribution resulting from the Markov chain induced by a given policy. Since such distributions necessitate recurrence, we propose a solution which ﬁnds policies that induce recurrent Markov chains within possibly non-recurrent Markov Decision Processes (MDPs). The SSPS problem functions as a generalization of steady-state control, which has been shown to be in PSPACE. We improve upon this result by showing that SSPS is in P via linear programming. Our results are validated using CPLEX simulations on MDPs with over 10000 states. We also prove that the deterministic variant of SSPS is NP-hard.

Journal of Artificial Intelligence Research, 2021

The planning domain has experienced increased interest in the formal synthesis of decision-making... more The planning domain has experienced increased interest in the formal synthesis of decision-making policies. This formal synthesis typically entails finding a policy which satisfies formal specifications in the form of some well-defined logic. While many such logics have been proposed with varying degrees of expressiveness and complexity in their capacity to capture desirable agent behavior, their value is limited when deriving decision-making policies which satisfy certain types of asymptotic behavior in general system models. In particular, we are interested in specifying constraints on the steady-state behavior of an agent, which captures the proportion of time an agent spends in each state as it interacts for an indefinite period of time with its environment. This is sometimes called the average or expected behavior of the agent and the associated planning problem is faced with significant challenges unless strong restrictions are imposed on the underlying model in terms of the c...

ArXiv, 2021

Deep neural networks such as AlphaFold and RoseTTAFold predict remarkably accurate structures of ... more Deep neural networks such as AlphaFold and RoseTTAFold predict remarkably accurate structures of proteins compared to other algorithmic approaches. It is known that biologically small perturbations in the protein sequence do not lead to drastic changes in the protein structure. In this paper, we demonstrate that RoseTTAFold does not exhibit such a robustness despite its high accuracy, and biologically small perturbations for some input sequences result in radically different predicted protein structures. This raises the challenge of detecting when these predicted protein structures cannot be trusted. We define the robustness measure for the predicted structure of a protein sequence to be the inverse of the root-mean-square distance (RMSD) in the predicted structure and the structure of its adversarially perturbed sequence. We use adversarial attack methods to create adversarial protein sequences, and show that the RMSD in the predicted protein structure ranges from 0.119Å to 34.162Å...

ArXiv, 2020

We propose a hardware learning rule for unsupervised clustering within a novel spintronic computi... more We propose a hardware learning rule for unsupervised clustering within a novel spintronic computing architecture. The proposed approach leverages the three-terminal structure of domain-wall magnetic tunnel junction devices to establish a feedback loop that serves to train such devices when they are used as synapses in a neuromorphic computing architecture.

Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, 2021

Several methods have recently been developed for computing attributions of a neural network's... more Several methods have recently been developed for computing attributions of a neural network's prediction over the input features. However, these existing approaches for computing attributions are noisy and not robust to small perturbations of the input. This paper uses the recently identified connection between dynamical systems and residual neural networks to show that the attributions computed over neural stochastic differential equations (SDEs) are less noisy, visually sharper, and quantitatively more robust. Using dynamical systems theory, we theoretically analyze the robustness of these attributions. We also experimentally demonstrate the efficacy of our approach in providing smoother, visually sharper and quantitatively robust attributions by computing attributions for ImageNet images using ResNet-50, WideResNet-101 models and ResNeXt-101 models.

ArXiv, 2021

The design of additive imperceptible perturbations to the inputs of deep classifiers to maximize ... more The design of additive imperceptible perturbations to the inputs of deep classifiers to maximize their misclassification rates is a central focus of adversarial machine learning. An alternative approach is to synthesize adversarial examples from scratch using GAN-like structures, albeit with the use of large amounts of training data. By contrast, this paper considers oneshot synthesis of adversarial examples; the inputs are synthesized from scratch to induce arbitrary soft predictions at the output of pre-trained models, while simultaneously maintaining high similarity to specified inputs. To this end, we present a problem that encodes objectives on the distance between the desired and output distributions of the trained model and the similarity between such inputs and the synthesized examples. We prove that the formulated problem is NP-complete. Then, we advance a generative approach to the solution in which the adversarial examples are obtained as the output of a generative networ...

Proceedings of the 30th on Symposium on Parallelism in Algorithms and Architectures, 2018

Algorithmic Aspects of Cloud Computing, 2019

Database migration is an ubiquitous need faced by enterprises that generate and use vast amount o... more Database migration is an ubiquitous need faced by enterprises that generate and use vast amount of data. This is due to database software updates, or from changes to hardware, project standards, and other business factors [1]. Migrating a large collection of databases is a way more challenging task than migrating a single database, due to the presence of additional constraints. These constraints include capacities of shifts, sizes of databases, and timing relationships. In this paper, we present a comprehensive framework that can be used to model database migration problems of different enterprises with customized constraints, by appropriately instantiating the parameters of the framework. We establish the computational complexities of a number of instantiations of this framework. We present fixed-parameter intractability results for various relevant parameters of the database migration problem. Finally, we discuss a randomized approximation algorithm for an interesting instantiation.

ArXiv, 2020

The planning domain has experienced increased interest in the formal synthesis of decision-making... more The planning domain has experienced increased interest in the formal synthesis of decision-making policies. This formal synthesis typically entails finding a policy which satisfies formal specifications in the form of some well-defined logic, such as Linear Temporal Logic (LTL) or Computation Tree Logic (CTL), among others. While such logics are very powerful and expressive in their capacity to capture desirable agent behavior, their value is limited when deriving decision-making policies which satisfy certain types of asymptotic behavior. In particular, we are interested in specifying constraints on the steady-state behavior of an agent, which captures the proportion of time an agent spends in each state as it interacts for an indefinite period of time with its environment. This is sometimes called the average or expected behavior of the agent. In this paper, we explore the steady-state planning problem of deriving a decision-making policy for an agent such that constraints on its ...

Memristor-based nano-crossbar computing is a revolutionary computing paradigm that does away with... more Memristor-based nano-crossbar computing is a revolutionary computing paradigm that does away with the traditional Von Neumann architectural separation of memory and computation units. The computation of Boolean formulas using memristor circuits has been a subject of several recent investigations. Crossbar computing, in general, has also been a topic of active interest, but sneak paths have posed a hurdle in the design of pervasive general-purpose crossbar computing paradigms. In this paper, we demonstrate that sneak paths in nano-crossbar computing can be exploited to design a Boolean-formula evaluation strategy. We demonstrate our approach on a simple Boolean formula and a 1-bit addition circuit. We also conjecture that our nano-crossbar design will be an effective approach for synthesizing high-performance customized arithmetic and logic circuits.

Proposed for presentation at the Conference on Magnetism and Magnetic Materials held November 2-6, 2020 in Virtual., 2020

International Joint Conference on Artificial Intelligence, 2020

2021 60th IEEE Conference on Decision and Control (CDC), 2021

2018 IEEE 36th International Conference on Computer Design (ICCD), 2018

2017 IEEE International Symposium on Circuits and Systems (ISCAS), 2017

International Joint Conference on Artificial Intelligence, 2019

2018 Design, Automation & Test in Europe Conference & Exhibition (DATE), 2018

Lecture Notes in Computer Science, 2018

Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, 2019

Journal of Artificial Intelligence Research, 2021

ArXiv, 2021

ArXiv, 2020

Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, 2021

ArXiv, 2021

Proceedings of the 30th on Symposium on Parallelism in Algorithms and Architectures, 2018

Algorithmic Aspects of Cloud Computing, 2019

ArXiv, 2020

The planning domain has experienced increased interest in the formal synthesis of decision-making... more The planning domain has experienced increased interest in the formal synthesis of decision-making policies. This formal synthesis typically entails finding a policy which satisfies formal specifications in the form of some well-defined logic, such as Linear Temporal Logic (LTL) or Computation Tree Logic (CTL), among others. While such logics are very powerful and expressive in their capacity to capture desirable agent behavior, their value is limited when deriving decision-making policies which satisfy certain types of asymptotic behavior. In particular, we are interested in specifying constraints on the steady-state behavior of an agent, which captures the proportion of time an agent spends in each state as it interacts for an indefinite period of time with its environment. This is sometimes called the average or expected behavior of the agent. In this paper, we explore the steady-state planning problem of deriving a decision-making policy for an agent such that constraints on its ...