arXiv subjects

Michael Chen

Publications and source records attributed to Michael Chen.

35 records · Page 2Linked to original sources

arXiv2024

The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning

The White House Executive Order on Artificial Intelligence highlights the risks of large language models (LLMs) empowering malicious actors in developing biological, cyber, and chemical weapons. To measure these risks of malicious use, government institutions and major AI labs are developing evaluations for hazardous capabilities in LLMs. However, current evaluations are private, preventing further research into mitigating risk. Furthermore, they focus on only a few, highly specific pathways for malicious use. To fill these gaps, we publicly release the Weapons of Mass Destruction Proxy (WMDP) benchmark, a dataset of 3,668 multiple-choice questions that serve as a proxy measurement of hazardous knowledge in biosecurity, cybersecurity, and chemical security. WMDP was developed by a consortium of academics and technical consultants, and was stringently filtered to eliminate sensitive information prior to public release. WMDP serves two roles: first, as an evaluation for hazardous knowledge in LLMs, and second, as a benchmark for unlearning methods to remove such hazardous knowledge. To guide progress on unlearning, we develop RMU, a state-of-the-art unlearning method based on controlling model representations. RMU reduces model performance on WMDP while maintaining general capabilities in areas such as biology and computer science, suggesting that unlearning may be a concrete path towards reducing malicious use from LLMs. We release our benchmark and code publicly at https://wmdp.ai

cs.LG↗

arXiv2023

AI Deception: A Survey of Examples, Risks, and Potential Solutions

This paper argues that a range of current AI systems have learned how to deceive humans. We define deception as the systematic inducement of false beliefs in the pursuit of some outcome other than the truth. We first survey empirical examples of AI deception, discussing both special-use AI systems (including Meta's CICERO) built for specific competitive situations, and general-purpose AI systems (such as large language models). Next, we detail several risks from AI deception, such as fraud, election tampering, and losing control of AI systems. Finally, we outline several potential solutions to the problems posed by AI deception: first, regulatory frameworks should subject AI systems that are capable of deception to robust risk-assessment requirements; second, policymakers should implement bot-or-not laws; and finally, policymakers should prioritize the funding of relevant research, including tools to detect AI deception and to make AI systems less deceptive. Policymakers, researchers, and the broader public should work proactively to prevent AI deception from destabilizing the shared foundations of our society.

cs.CY↗

arXiv2023

Comment on Transverse Charge Density and the Radius of the Proton

The charge radius of the proton is typically determined from electron-proton scattering by extracting the proton's electric form factor and then making use of the derivative of that form factor at zero four-momentum transfer. Unfortunately, experimentally, one cannot measure to zero four-momentum transfer and thus extrapolation is often required. In the work of Alexander Gramolin and Rebecca Russell, they present a novel method that does not use the slope and found a radius of the proton that contradicts many other recent results. Our analysis of their paper discusses some issues with this method and we show that by simply changing the binning of the data and/or including an additional set of data the results change dramatically.

nucl-ex↗

arXiv2020

Deep Phase Decoder: Self-calibrating phase microscopy with an untrained deep neural network

Deep neural networks have emerged as effective tools for computational imaging including quantitative phase microscopy of transparent samples. To reconstruct phase from intensity, current approaches rely on supervised learning with training examples; consequently, their performance is sensitive to a match of training and imaging settings. Here we propose a new approach to phase microscopy by using an untrained deep neural network for measurement formation, encapsulating the image prior and imaging physics. Our approach does not require any training data and simultaneously reconstructs the sought phase and pupil-plane aberrations by fitting the weights of the network to the captured images. To demonstrate experimentally, we reconstruct quantitative phase from through-focus images blindly (i.e. no explicit knowledge of the aberrations).

eess.IV↗

arXiv2019

High-resolution 3D refractive index microscopy of multiple-scattering samples from intensity images

Optical diffraction tomography (ODT) reconstructs a samples volumetric refractive index (RI) to create high-contrast, quantitative 3D visualizations of biological samples. However, standard implementations of ODT use interferometric systems, and so are sensitive to phase instabilities, complex mechanical design, and coherent noise. Furthermore, their reconstruction framework is typically limited to weakly-scattering samples, and thus excludes a whole class of multiple-scattering samples. Here, we implement a new 3D RI microscopy technique that utilizes a computational multi-slice beam propagation method to invert the optical scattering process and reconstruct high-resolution (NA>1.0) 3D RI distributions of multiple-scattering samples. The method acquires intensity-only measurements from different illumination angles, and then solves a non-linear optimization problem to recover the sample 3D RI distribution. We experimentally demonstrate reconstruction of samples with varying amounts of multiple scattering: a 3T3 fibroblast cell, a cluster of C. elegans embryos, and a whole C. elegans worm, with lateral and axial resolutions of 250 nm and 900 nm, respectively.

eess.IV↗

arXiv2019

CODAH: An Adversarially Authored Question-Answer Dataset for Common Sense

Commonsense reasoning is a critical AI capability, but it is difficult to construct challenging datasets that test common sense. Recent neural question answering systems, based on large pre-trained models of language, have already achieved near-human-level performance on commonsense knowledge benchmarks. These systems do not possess human-level common sense, but are able to exploit limitations of the datasets to achieve human-level scores. We introduce the CODAH dataset, an adversarially-constructed evaluation dataset for testing common sense. CODAH forms a challenging extension to the recently-proposed SWAG dataset, which tests commonsense knowledge using sentence-completion questions that describe situations observed in video. To produce a more difficult dataset, we introduce a novel procedure for question acquisition in which workers author questions designed to target weaknesses of state-of-the-art neural question answering systems. Workers are rewarded for submissions that models fail to answer correctly both before and after fine-tuning (in cross-validation). We create 2.8k questions via this procedure and evaluate the performance of multiple state-of-the-art question answering systems on our dataset. We observe a significant gap between human performance, which is 95.3%, and the performance of the best baseline accuracy of 67.5% by the BERT-Large model.

cs.CL↗

arXiv2019

Data-Driven Design for Fourier Ptychographic Microscopy

Fourier Ptychographic Microscopy (FPM) is a computational imaging method that is able to super-resolve features beyond the diffraction-limit set by the objective lens of a traditional microscope. This is accomplished by using synthetic aperture and phase retrieval algorithms to combine many measurements captured by an LED array microscope with programmable source patterns. FPM provides simultaneous large field-of-view and high resolution imaging, but at the cost of reduced temporal resolution, thereby limiting live cell applications. In this work, we learn LED source pattern designs that compress the many required measurements into only a few, with negligible loss in reconstruction quality or resolution. This is accomplished by recasting the super-resolution reconstruction as a Physics-based Neural Network and learning the experimental design to optimize the network's overall performance. Specifically, we learn LED patterns for different applications (e.g. amplitude contrast and quantitative phase imaging) and show that the designs we learn through simulation generalize well in the experimental setting. Further, we discuss a context-specific loss function, practical memory limitations, and interpretability of our learned designs.

eess.SP↗

arXiv2018

Theory, design, and experimental verification of a reflectionless bianisotropic Huygens' metasurface for wide-angle refraction

Huygens' metasurfaces are electrically thin devices which allow arbitrary field transformations. Beam refraction is among the first demonstrations of realized metasurfaces. As previously shown for extreme-angle refraction, control over only the electric impedance and magnetic admittance of the Huygens' metasurface proved insufficient to produce the desired reflectionless field transformation. To maintain zero reflections for wide refraction angles, magnetoelectric coupling between the electric and magnetic response of the metasurface, leading to bianisotropy, can be introduced. In this paper, we report the theory, design, and experimental characterization of a reflectionless bianisotropic metasurface for extreme-angle refraction of a normally incident plane wave towards 71.8$^\circ$ at 20 GHz. The theory and design of three-layer asymmetric bianisotropic unit cells are discussed. The realized printed circuit board structure was tested via full-wave simulations as well as experimental characterization. To experimentally verify the prototype, two setups were used. A quasi-optical experiment was conducted to assess the specular reflections of the metasurface, while a far-field antenna measurement characterized its refraction nature. The measurements verify that the fabricated metasurface has negligible reflections and the majority of the scattered power is refracted to the desired Floquet mode. This provides an experimental demonstration of a reflectionless wide-angle refracting metasurface using a bianisotropic Huygens' metasurface at microwave frequencies.

physics.app-ph↗

arXiv2018

A Leaky-Wave Antenna With Controlled Radiation Using a Bianisotropic Huygens' Metasurface

In this paper, a novel concept of a leaky-wave antenna is proposed, based on the use of Huygens' metasurfaces. It consists of a parallel-plate waveguide in which the top plate is replaced by a bianisotropic metasurface of the Omega type. It is shown that there is an exact solution to transform the guided mode into a leaky-mode with arbitrary control of the constant leakage factor and the pointing direction. Although the solution turns out to be periodic, only one Floquet mode is excited and radiates, even for electrically long periods. Thanks to the intrinsic spurious Floquet mode suppression, broadside radiation can be achieved without any degradation. Simulations with idealized reactance sheets verify the concept. Moreover, physical structures compatible with PCB fabrication have been proposed and designed, considering aspects such as the effect of losses. Finally, experimental results of two prototypes are presented and discussed.

physics.app-ph↗

arXiv2018

A Practical Reconstruction Method for Three-Dimensional Phase Contrast Atomic Electron Tomography

Electron tomography is a technique used in both materials science and structural biology to image features well below optical resolution limit. In this work, we present a new algorithm for reconstructing the three-dimensional(3D) electrostatic potential of a sample at atomic resolution from phase contrast imaging using high-resolution transmission electron microscopy. Our method accounts for dynamical and strong phase scattering, providing more accurate results with much lower electron doses than those current atomic electron tomography experiments. We test our algorithm using simulated images of a synthetic needle geometry dataset composed of an amorphous silicon dioxide shell around a silicon core. Our results show that, for a wide range of experimental parameters, we can accurately determine both atomic positions and species, and also identify vacancies even for light elements such as silicon and disordered materials such as amorphous silicon dioxide and also identify vacancies.

eess.SP↗

arXiv2018

Drive2Vec: Multiscale State-Space Embedding of Vehicular Sensor Data

With automobiles becoming increasingly reliant on sensors to perform various driving tasks, it is important to encode the relevant CAN bus sensor data in a way that captures the general state of the vehicle in a compact form. In this paper, we develop a deep learning-based method, called Drive2Vec, for embedding such sensor data in a low-dimensional yet actionable form. Our method is based on stacked gated recurrent units (GRUs). It accepts a short interval of automobile sensor data as input and computes a low-dimensional representation of that data, which can then be used to accurately solve a range of tasks. With this representation, we (1) predict the exact values of the sensors in the short term (up to three seconds in the future), (2) forecast the long-term average values of these same sensors, (3) infer additional contextual information that is not encoded in the data, including the identity of the driver behind the wheel, and (4) build a knowledge base that can be used to auto-label data and identify risky states. We evaluate our approach on a dataset collected by Audi, which equipped a fleet of test vehicles with data loggers to store all sensor readings on 2,098 hours of driving on real roads. We show in several experiments that our method outperforms other baselines by up to 90%, and we further demonstrate how these embeddings of sensor data can be used to solve a variety of real-world automotive applications.

cs.LG↗

arXiv2018

Best arm identification in multi-armed bandits with delayed feedback

We propose a generalization of the best arm identification problem in stochastic multi-armed bandits (MAB) to the setting where every pull of an arm is associated with delayed feedback. The delay in feedback increases the effective sample complexity of standard algorithms, but can be offset if we have access to partial feedback received before a pull is completed. We propose a general framework to model the relationship between partial and delayed feedback, and as a special case we introduce efficient algorithms for settings where the partial feedback are biased or unbiased estimators of the delayed feedback. Additionally, we propose a novel extension of the algorithms to the parallel MAB setting where an agent can control a batch of arms. Our experiments in real-world settings, involving policy search and hyperparameter optimization in computational sustainability domains for fast charging of batteries and wildlife corridor construction, demonstrate that exploiting the structure of partial feedback can lead to significant improvements over baselines in both sequential and parallel MAB.

cs.LG↗

arXiv2017

UNBIAS PUF: A Physical Implementation Bias Agnostic Strong PUF

The Physical Unclonable Function (PUF) is a promising hardware security primitive because of its inherent uniqueness and low cost. To extract the device-specific variation from delay-based strong PUFs, complex routing constraints are imposed to achieve symmetric path delays; and systematic variations can severely compromise the uniqueness of the PUF. In addition, the metastability of the arbiter circuit of an Arbiter PUF can also degrade the quality of the PUF due to the induced instability. In this paper we propose a novel strong UNBIAS PUF that can be implemented purely by Register Transfer Language (RTL), such as verilog, without imposing any physical design constraints or delay characterization effort to solve the aforementioned issues. Efficient inspection bit prediction models for unbiased response extraction are proposed and validated. Our experimental results of the strong UNBIAS PUF show 5.9% intra-Fractional Hamming Distance (FHD) and 45.1% inter-FHD on 7 Field Programmable Gate Array (FPGA) boards without applying any physical layout constraints or additional XOR gates. The UNBIAS PUF is also scalable because no characterization cost is required for each challenge to compensate the implementation bias. The averaged intra-FHD measured at worst temperature and voltage variation conditions is 12%, which is still below the margin of practical Error Correction Code (ECC) with error reduction techniques for PUFs.

cs.AR↗

arXiv2017

Experimental Verification of Reflectionless Wide-Angle Refraction via a Bianisotropic Huygens' Metasurface

We report the design, fabrication, and characterization of bianisotropic Huygens' metasurfaces (BHMSs) for refraction of normally incident beams towards 71.8 degrees. As previously shown, all three BHMS degrees of freedom, namely, electric polarizability, magnetic polarizability and omega-type magnetoelectric coupling, are required to ensure no reflections occur for such wide-angle impedance mismatch. The unit cells are composed of three metallic layers, yielding a printed-circuit-board (PCB) structure. The fabricated BHMS is characterized in a quasi-optical setup, used to accurately assess specular reflections. Subsequently, the horn-illuminated BHMS' radiation pattern is measured in a far-field chamber, to evaluate the device's refraction characteristics. The measured results verify that the BHMS has negligible reflections, and the majority of the scattered power is coupled to the desirable Floquet-Bloch mode. To the best of our knowledge, this is the first experimental demonstration of such a reflectionless wide-angle refracting metasurface.

physics.optics↗

arXiv2015

Experimental robustness of Fourier Ptychography phase retrieval algorithms

Fourier ptychography is a new computational microscopy technique that provides gigapixel-scale intensity and phase images with both wide field-of-view and high resolution. By capturing a stack of low-resolution images under different illumination angles, a nonlinear inverse algorithm can be used to computationally reconstruct the high-resolution complex field. Here, we compare and classify multiple proposed inverse algorithms in terms of experimental robustness. We find that the main sources of error are noise, aberrations and mis-calibration (i.e. model mis-match). Using simulations and experiments, we demonstrate that the choice of cost function plays a critical role, with amplitude-based cost functions performing better than intensity-based ones. The reason for this is that Fourier ptychography datasets consist of images from both brightfield and darkfield illumination, representing a large range of measured intensities. Both noise (e.g. Poisson noise) and model mis-match errors are shown to scale with intensity. Hence, algorithms that use an appropriate cost function will be more tolerant to both noise and model mis-match. Given these insights, we propose a global Newton's method algorithm which is robust and computationally efficient. Finally, we discuss the impact of procedures for algorithmic correction of aberrations and mis-calibration.

physics.optics↗

arXiv2015

Computational illumination for high-speed in vitro Fourier ptychographic microscopy

We demonstrate a new computational illumination technique that achieves large space-bandwidth-time product, for quantitative phase imaging of unstained live samples in vitro. Microscope lenses can have either large field of view (FOV) or high resolution, not both. Fourier ptychographic microscopy (FPM) is a new computational imaging technique that circumvents this limit by fusing information from multiple images taken with different illumination angles. The result is a gigapixel-scale image having both wide FOV and high resolution, i.e. large space-bandwidth product (SBP). FPM has enormous potential for revolutionizing microscopy and has already found application in digital pathology. However, it suffers from long acquisition times (on the order of minutes), limiting throughput. Faster capture times would not only improve imaging speed, but also allow studies of live samples, where motion artifacts degrade results. In contrast to fixed (e.g. pathology) slides, live samples are continuously evolving at various spatial and temporal scales. Here, we present a new source coding scheme, along with real-time hardware control, to achieve 0.8 NA resolution across a 4x FOV with sub-second capture times. We propose an improved algorithm and new initialization scheme, which allow robust phase reconstruction over long time-lapse experiments. We present the first FPM results for both growing and confluent in vitro cell cultures, capturing videos of subcellular dynamical phenomena in popular cell lines undergoing division and migration. Our method opens up FPM to applications with live samples, for observing rare events in both space and time.

physics.optics↗

arXiv2012

Radial elasticity of multi-walled carbon nanotubes

We report an experimental and a theoretical study of the radial elasticity of multi-walled carbon nanotubes as a function of external radius. We use atomic force microscopy and apply small indentation amplitudes in order to stay in the linear elasticity regime. The number of layers for a given tube radius is inferred from transmission electron microscopy, revealing constant ratios of external to internal radii. This enables a comparison with molecular dynamics results, which also shed some light onto the applicability of Hertz theory in this context. Using this theory, we find a radial Young modulus strongly decreasing with increasing radius and reaching an asymptotic value of 30 +/- 10 GPa.

cond-mat.mtrl-sci↗