Source author record

James Smith

James Smith appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Machine Learning Computer Vision Logic in Computer Science Artificial Intelligence cond-mat.mtrl-sci Cryptography and Security math.AP math.FA Networking and Internet Architecture physics.app-ph physics.chem-ph Programming Languages Social and Information Networks Software Engineering

Catalog footprint

What is connected

8works

14topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

A Closer Look at Knowledge Distillation with Features, Logits, and Gradients

Knowledge distillation (KD) is a substantial strategy for transferring learned knowledge from one neural network model to another. A vast number of methods have been developed for this strategy. While most method designs a more efficient way to facilitate knowledge transfer, less attention has been put on comparing the effect of knowledge sources such as features, logits, and gradients. This work provides a new perspective to motivate a set of knowledge distillation strategies by approximating the classical KL-divergence criteria with different knowledge sources, making a systematic comparison possible in model compression and incremental learning. Our analysis indicates that logits are generally a more efficient knowledge source and suggests that having sufficient feature dimensions is crucial for the model design, providing a practical guideline for effective KD-based transfer learning.

preprint2022arXiv

Lifelong Wandering: A realistic few-shot online continual learning setting

Online few-shot learning describes a setting where models are trained and evaluated on a stream of data while learning emerging classes. While prior work in this setting has achieved very promising performance on instance classification when learning from data-streams composed of a single indoor environment, we propose to extend this setting to consider object classification on a series of several indoor environments, which is likely to occur in applications such as robotics. Importantly, our setting, which we refer to as online few-shot continual learning, injects the well-studied issue of catastrophic forgetting into the few-shot online learning paradigm. In this work, we benchmark several existing methods and adapted baselines within our setting, and show there exists a trade-off between catastrophic forgetting and online performance. Our findings motivate the need for future work in this setting, which can achieve better online performance without catastrophic forgetting.

preprint2021arXiv

On the Adversarial Robustness of Quantized Neural Networks

Reducing the size of neural network models is a critical step in moving AI from a cloud-centric to an edge-centric (i.e. on-device) compute paradigm. This shift from cloud to edge is motivated by a number of factors including reduced latency, improved security, and higher flexibility of AI algorithms across several application domains (e.g. transportation, healthcare, defense, etc.). However, it is currently unclear how model compression techniques may affect the robustness of AI algorithms against adversarial attacks. This paper explores the effect of quantization, one of the most common compression techniques, on the adversarial robustness of neural networks. Specifically, we investigate and model the accuracy of quantized neural networks on adversarially-perturbed images. Results indicate that for simple gradient-based attacks, quantization can either improve or degrade adversarial robustness depending on the attack strength.

preprint2020arXiv

Pulsed laser deposition of single phase n- and p-type Cu2O thin films with low resistivity

Low resistivity (~3-24 mOhm.cm) with tunable n- and p-type phase pure Cu2O thin films have been grown by pulsed laser deposition at 25-200 0C by varying the background oxygen partial pressure (O2pp). Capacitance data obtained by electrochemical impedance spectroscopy was used to determine the conductivity (n- or p-type), carrier density, and flat band potentials for samples grown on indium tin oxide (ITO) at 25 0C. The Hall mobility of the n- and p-type Cu2O was estimated to be ~ 0.85 cm2.V-1s-1 and ~ 4.78 cm2.V-1s-1 respectively for samples grown on quartz substrate at 25 0C. An elevated substrate temperature ~ 200 0C with O2pp = 2 - 3 mTorr yielded p-type Cu2O films with six orders of magnitude higher resistivities in the range ~ 9 - 49 kOhm.cm and mobilities in the range ~ 13.5 - 22.2 cm2.V-1s-1. UV-Vis-NIR diffuse reflectance spectroscopy showed optical bandgaps of Cu2O films in the range of 1.76 to 2.15 eV depending on O2pp. Thin films grown at oxygen-rich conditions O2pp > 7 mTorr yielded mixed-phase copper oxide irrespective of the substrate temperatures and upon air annealing at 550 0C for 1 hour completely converted to CuO phase with n-type semiconducting properties (~12 Ohm.cm, ~1.50 cm2.V-1s-1). The as-grown p- and n-type Cu2O showed rectification and a photovoltaic (PV) response in solid junctions with n-ZnO and p-Si electrodes respectively. Our findings may create new opportunities for devising Cu2O based junctions requiring low process temperatures.

preprint2014arXiv

Gender Prediction in Social Media

In this paper, we explore the task of gender classification using limited network data with an application to Fotolog. We take a heuristic approach to automating gender inference based on username, followers and network structure. We test our approach on a subset of 100,000 nodes and analyze our results to find that there is a lot of value in these limited information and that there is great promise in further pursuing this approach to classification.

preprint2013arXiv

Proving Properties of Rich Internet Applications

We introduce application layer specifications, which allow us to reason about the state and transactions of rich Internet applications. We define variants of the state/event based logic UCTL* along with two example applications to demonstrate this approach, and then look at a distributed, rich Internet application, proving properties about the information it stores and disseminates. Our approach enables us to justify proofs about abstract properties that are preserved in the face of concurrent, networked inputs by proofs about concrete properties in an Internet setting. We conclude that our approach makes it possible to reason about the programs and protocols that comprise the Internet's application layer with reliability and generality.

preprint2012arXiv

The Jasper Framework: Towards a Platform Independent, Formal Treatment of Web Programming

This paper introduces Jasper, a web programming framework which allows web applications to be developed in an essentially platform indepedent manner and which is also suited to a formal treatment. It outlines Jasper conceptually and shows how Jasper is implemented on several commonplace platforms. It also introduces the Jasper Music Store, a web application powered by Jasper and implemented on each of these platforms. And it briefly describes a formal treatment and outlines the tools and languages planned that will allow this treatment to be automated.

preprint2009arXiv

Dispersive and Strichartz estimates for hyperbolic equations with constant coefficients

Dispersive and Strichartz estimates for solutions to general strictly hyperbolic partial differential equations with constant coefficients are considered. The global time decay estimates of $L^p-L^q$ norms of propagators are obtained, and it is shown how the time decay rates depend on the geometry of the problem. The frequency space is separated in several zones each giving a certain decay rate. Geometric conditions on characteristics responsible for the particular decay are identified and investigated. Thus, a comprehensive analysis is carried out for strictly hyperbolic equations of high orders with lower order terms of a general form. Results are applied to establish time decay estimates for the Fokker-Planck equation and for semilinear hyperbolic equations.