arXiv subjects

Steven Lee

Publications and source records attributed to Steven Lee.

18 recordsLinked to original sources

arXiv2025

Optimal Pure Differentially Private Sparse Histograms in Deterministic Linear Time

We present an algorithm that releases a pure differentially private (under the replacement neighboring relation) sparse histogram for $n$ participants over a domain of size $d \gg n$. Our method achieves the optimal $\ell_\infty$-estimation error and runs in strictly $O(n)$ time in the Word-RAM model, improving upon the previous best deterministic-time bound of $\tilde{O}(n^2)$ and resolving the open problem of breaking this quadratic barrier (Balcer and Vadhan, 2019). Moreover, the algorithm admits an efficient circuit implementation, enabling the first near-linear communication and computation cost pure DP histogram MPC protocol with optimal $\ell_\infty$-estimation error. Central to our algorithm is a novel **private item blanket** technique with target-length padding, which hides differences in the supports of neighboring histograms while remaining efficiently implementable.

cs.DS↗

arXiv2025

Amorphous Fortress Online: Collaboratively Designing Open-Ended Multi-Agent AI and Game Environments

This work introduces Amorphous Fortress Online -- a web-based platform where users can design petri-dish-like environments and games consisting of multi-agent AI characters. Users can play, create, and share artificial life and game environments made up of microscopic but transparent finite-state machine agents that interact with each other. The website features multiple interactive editors and accessible settings to view the multi-agent interactions directly from the browser. This system serves to provide a database of thematically diverse AI and game environments that use the emergent behaviors of simple AI agents.

cs.AI↗

arXiv2024

Safe Autonomy for Uncrewed Surface Vehicles Using Adaptive Control and Reachability Analysis

Marine robots must maintain precise control and ensure safety during tasks like ocean monitoring, even when encountering unpredictable disturbances that affect performance. Designing algorithms for uncrewed surface vehicles (USVs) requires accounting for these disturbances to control the vehicle and ensure it avoids obstacles. While adaptive control has addressed USV control challenges, real-world applications are limited, and certifying USV safety amidst unexpected disturbances remains difficult. To tackle control issues, we employ a model reference adaptive controller (MRAC) to stabilize the USV along a desired trajectory. For safety certification, we developed a reachability module with a moving horizon estimator (MHE) to estimate disturbances affecting the USV. This estimate is propagated through a forward reachable set calculation, predicting future states and enabling real-time safety certification. We tested our safe autonomy pipeline on a Clearpath Heron USV in the Charles River, near MIT. Our experiments demonstrated that the USV's MRAC controller and reachability module could adapt to disturbances like thruster failures and drag forces. The MRAC controller outperformed a PID baseline, showing a 45%-81% reduction in RMSE position error. Additionally, the reachability module provided real-time safety certification, ensuring the USV's safety. We further validated our pipeline's effectiveness in underway replenishment and canal scenarios, simulating relevant marine tasks.

cs.RO↗

arXiv2023

Online Data-Driven Safety Certification for Systems Subject to Unknown Disturbances

Deploying autonomous systems in safety critical settings necessitates methods to verify their safety properties. This is challenging because real-world systems may be subject to disturbances that affect their performance, but are unknown a priori. This work develops a safety-verification strategy wherein data is collected online and incorporated into a reachability analysis approach to check in real-time that the system avoids dangerous regions of the state space. Specifically, we employ an optimization-based moving horizon estimator (MHE) to characterize the disturbance affecting the system, which is incorporated into an online reachability calculation. Reachable sets are calculated using a computational graph analysis tool to predict the possible future states of the system and verify that they satisfy safety constraints. We include theoretical arguments proving our approach generates reachable sets that bound the future states of the system, as well as numerical results demonstrating how it can be used for safety verification. Finally, we present results from hardware experiments demonstrating our approach's ability to perform online reachability calculations for an unmanned surface vehicle subject to currents and actuator failures.

eess.SY↗

arXiv2023

Tracing the colliding winds of $\eta$ Carinae in He I

$\eta$ Carinae is an extremely luminous and energetic colliding-wind binary. The combination of its orbit and orientation, with respect to our line of sight, enables direct investigation of the conditions and geometry of the colliding winds. We analyse optical He I 5876 and 7065 $\unicode{x212B}$ line profiles from the Global Jet Watch observatories covering the last 1.3 orbital periods. The sustained coverage throughout apastron reveals the distinct dynamics of the emitting versus absorbing components: the emission lines follow orbital velocities whilst one of the absorption lines is detected only around apastron ($0.08 < \phi < 0.95$) and exhibits velocities that deviate substantially from the orbital motion. To interpret these deviations, we conjecture that this He I absorption component is formed in the post-shock primary wind, and is only detected when our line of sight intersects with the shock cone formed by the collision of the two winds. We formulate a geometrical model for the colliding winds in terms of a hyperboloid in which the opening angle and location of its apex are parameterised in terms of the ratio of the wind momentum of the primary star to that of companion. We fit this geometrical model to the absorption velocities, finding results that are concordant with the panchromatic observations and simulations of $\eta$ Carinae. The model presented here is an extremely sensitive probe of the exact geometry of the wind momentum balance of binary stars, and can be extended to probe the latitudinal dependence of mass loss.

astro-ph.SR↗

arXiv2021

The circumbinary rings of GG Carinae: indications of disc eccentricity growth in the B[e] supergiant's atomic emission lines

B[e] supergiants have unusual circumstellar environments which may include thin, concentric rings displaying atomic line emission. GG Carinae is a B[e] supergiant binary which exhibits such a geometry in its circumbinary environment. We study atomic emission lines arising from GG Carinae's circumbinary disc in FEROS spectra collected between 1998 and 2015. We find that semi-forbidden Fe\,II] and permitted Ca\,II emission are formed in the same thin circumbinary ring previously reported to have forbidden [O\,I] and [Ca\,II] emission. We find that there are two circumbinary rings orbiting with projected velocities of $84.6\pm1.0$\,km\,s$^{-1}$ and $27.3\pm0.6$\,km\,s$^{-1}$. Deprojecting these velocities from the line-of-sight, and using updated binary masses presented by \cite{Porter2021GGPhotometry}, we find that the radii of the circumbinary rings are $2.8^{+0.9}_{-1.1}$\,AU and $27^{+9}_{-10}$\,AU for the inner ring and outer ring respectively. We find evidence of subtle dynamical change in the inner circumbinary ring over the 17 years spanned by the data, manifesting in variability in the ratio of the intensity of the blueshifted peak to the redshifted peak of its emission lines and the central velocity becoming more blueshifted. We perform smoothed-particle hydrodynamic simulations of the system which suggest that these observed changes are consistent with pumping of the eccentricity of a radially thin circumbinary ring by the inner binary. We find a systemic velocity of the GG Carinae system of $-23.2 \pm 0.4$\,km\,s$^{-1}$.

astro-ph.SR↗

arXiv2021

The onset of jets in classical novae

We present two further classical novae, V906 Car and V5668 Sgr, that show jets and accretion disc spectral signatures in their H-alpha complexes throughout the first 1000 days following their eruptions. From extensive densely time-sampled spectroscopy, we measure the appearance of the first high-velocity absorption component in V906 Car, and the duration of the commencement of the main H-alpha emission. We constrain the time taken for V5668 Sgr to transition to the nebular phase using [N II] 6584\r{A}. We find these timings to be consistent with the jet and accretion disc model for explaining optical spectral line profile changes in classical novae, and discuss the implications of this model for enrichment of the interstellar medium.

astro-ph.HE↗

arXiv2021

GG Carinae: Discovery of orbital phase dependent 1.583-day periodicities in the B[e] supergiant binary

GG Carinae is a binary whose primary component is a B[e] supergiant. Using photometric data from TESS, ASAS, OMC, and ASAS-SN, and spectroscopic data from the Global Jet Watch to study visible He\,I, Fe\,II and Si\,II emission lines, we investigate the short-period variations which are exhibited in GG Car. We find a hitherto neglected periodicity of $1.583156\pm0.0002$\,days that is present in both its photometry and the radial velocities of its emission lines, alongside variability at the well-established $\sim$31-day orbital period. We find that the amplitudes of the shorter-period variations in both photometry and some of the emission lines are modulated by the orbital phase of the binary, such that the short-period variations have largest amplitudes when the binary is at periastron. There are no significant changes in the phases of the short-period variations over the orbital period. We investigate potential causes of the 1.583-day variability, and find that the observed period agrees well with the expected period of the $l=2$ f-mode of the primary given its mass and radius. We propose that the primary is periodically pulled out of hydrostatic equilibrium by the quadrupolar tidal forces when the components are near periastron in the binary's eccentric orbit ($e=0.5$) and the primary almost fills its Roche lobe. This causes an oscillation at the $l=2$ f-mode frequency which is damped as the distance between the components increases.

astro-ph.SR↗

arXiv2021

The precessing jets of classical nova YZ Reticuli

The classical nova YZ Reticuli was discovered in July 2020. Shortly after this we commenced a sustained, highly time-sampled coverage of its subsequent rapid evolution with time-resolved spectroscopy from the Global Jet Watch observatories. Its H-alpha complex exhibited qualitatively different spectral signatures in the following weeks and months. We find that these H-alpha complexes are well described by the same five Gaussian emission components throughout the six months following eruption. These five components appear to constitute two pairs of lines, from jet outflows and an accretion disc, together with an additional central component. The correlated, symmetric patterns that these jet/accretion disc pairs exhibit suggest precession, probably in response to the large perturbation caused by the nova eruption. The jet and accretion disc signatures persist from the first ten days after brightening -- evidence that the accretion disc survived the disruption. We also compare another classical nova (V6568 Sgr) that erupted in July 2020 whose H-alpha complex can be described analogously, but with faster line-of-sight jet speeds exceeding 4000 km/s. We suggest that classical novae with higher mass white dwarfs bridge the gap between recurrent novae and classical novae such as YZ Reticuli.

astro-ph.HE↗

arXiv2021

Playing with Food: Learning Food Item Representations through Interactive Exploration

A key challenge in robotic food manipulation is modeling the material properties of diverse and deformable food items. We propose using a multimodal sensory approach to interact and play with food that facilitates the ability to distinguish these properties across food items. First, we use a robotic arm and an array of sensors, which are synchronized using ROS, to collect a diverse dataset consisting of 21 unique food items with varying slices and properties. Afterwards, we learn visual embedding networks that utilize a combination of proprioceptive, audio, and visual data to encode similarities among food items using a triplet loss formulation. Our evaluations show that embeddings learned through interactions can successfully increase performance in a wide range of material and shape classification tasks. We envision that these learned embeddings can be utilized as a basis for planning and selecting optimal parameters for more material-aware robotic food manipulation skills. Furthermore, we hope to stimulate further innovations in the field of food robotics by sharing this food playing dataset with the research community.

cs.RO↗

arXiv2020

GG Carinae: orbital parameters and accretion indicators from phase-resolved spectroscopy and photometry

B[\,e\,] supergiants are a rare and unusual class of massive and luminous stars, characterised by opaque circumstellar envelopes. GG Carinae is a binary whose primary component is a B[\,e\,] supergiant and whose variability has remained unsatisfactorily explained. Using photometric data from ASAS, OMC, and ASAS-SN, and spectroscopic data from the Global Jet Watch and FEROS to study visible emission lines, we focus on the variability of the system at its $\sim$31-day orbital period and constrain the stellar parameters of the primary. There is one photometric minimum per orbital period and, in the emission line spectroscopy, we find a correlation between the amplitude of radial velocity variations and the initial energy of the line species. The spectral behaviour is consistent with the emission lines forming in the primary's wind, with the variable amplitudes between line species being caused by the less energetic lines forming at larger radii on average. By modelling the atmosphere of the primary, we are able to model the radial velocity variations of the wind lines in order to constrain the orbit of the binary. We find that the binary is even more eccentric than previously believed ($e=0.5\pm0.03$). Using this orbital solution, the system is brightest at periastron and dimmest at apastron, and the shape of the photometric variations at the orbital period can be well described by the variable accretion by the secondary of the primary's wind. We suggest that the evolutionary history of GG Carinae may need to be reevaluated in a binary context.

astro-ph.SR↗

arXiv2020

Learning Skills to Patch Plans Based on Inaccurate Models

Planners using accurate models can be effective for accomplishing manipulation tasks in the real world, but are typically highly specialized and require significant fine-tuning to be reliable. Meanwhile, learning is useful for adaptation, but can require a substantial amount of data collection. In this paper, we propose a method that improves the efficiency of sub-optimal planners with approximate but simple and fast models by switching to a model-free policy when unexpected transitions are observed. Unlike previous work, our method specifically addresses when the planner fails due to transition model error by patching with a local policy only where needed. First, we use a sub-optimal model-based planner to perform a task until model failure is detected. Next, we learn a local model-free policy from expert demonstrations to complete the task in regions where the model failed. To show the efficacy of our method, we perform experiments with a shape insertion puzzle and compare our results to both pure planning and imitation learning approaches. We then apply our method to a door opening task. Our experiments demonstrate that our patch-enhanced planner performs more reliably than pure planning and with lower overall sample complexity than pure imitation learning.

cs.RO↗

arXiv2020

Classical Nova Carinae 2018: Discovery of circumbinary iron and oxygen

We present time-lapse spectroscopy of a classical nova explosion commencing 9 days after discovery. These data reveal the appearance of a transient feature in Fe II and [O I]. We explore different models for this feature and conclude that it is best explained by a circumbinary disc shock-heated following the classical nova event. Circumbinary discs may play an important role in novae in accounting for the absorption systems known as THEA, the transfer of angular momentum, and the possible triggering of the nova event itself.

astro-ph.HE↗

arXiv2019

The Huntsman Telescope

The Huntsman Telescope, located at Siding Spring Observatory in Australia, is a system of ten telephoto Canon lenses designed for low surface brightness imaging in the Southern sky. Based upon the Dragonfly Telephoto Array, the refractive lens-based system provides an obstruction free optical path, which reduces the number of scattering surfaces and allows easier access to lower surface brightness levels. In this proceeding, we present an analysis of the impact of flat fielding uncertainty on the limiting low surface brightness levels. We show that a fairly standard set of flat-field data can be well-characterised to a $\sim0.1\%$ level. This corresponds to a 5-$\sigma$ lower limit of $\sim33$ magnitude per arcsecond$^2$, which means that flat fielding is not likely going to set Huntsman's low surface brightness limit. We also present early results of an exoplanet transient mode for Huntsman where all lenses work together to detect subtle variations in the luminosity of relatively bright $V=8-12$ magnitude stars. High-precision exoplanet imaging is ultimately limited by systematic uncertainties, so we anticipate multiple lenses will help to mitigate issues related to pixel-to-pixel and intra-pixel sensitivity variations. Our initial results show we can easily get $\sim0.4\%$ photometric precision with a single, defocused lens.

astro-ph.IM↗

arXiv2018

SS433's jet trace from ALMA imaging and Global Jet Watch spectroscopy: evidence for post-launch particle acceleration

We present a comparison of Doppler-shifted H-alpha line emission observed by the Global Jet Watch from freshly-launched jet ejecta at the nucleus of the Galactic microquasar SS433 with subsequent ALMA imaging at mm-wavelengths of the same jet ejecta. There is a remarkable similarity between the transversely-resolved synchrotron emission and the prediction of the jet trace from optical spectroscopy: this is an a priori prediction not an a posteriori fit, confirming the ballistic nature of the jet propagation. The mm-wavelength of the ALMA polarimetry is sufficiently short that the Faraday rotation is negligible and therefore that the observed E-vector directions are accurately orthogonal to the projected local magnetic field. Close to the nucleus the B-field vectors are perpendicular to the direction of propagation. Further out from the nucleus, the B-field vectors that are coincident with the jet instead become parallel to the ridge line; this occurs at a distance where the jet bolides are expected to expand into one another. X-ray variability has also been observed at this location; this has a natural explanation if shocks from the expanding and colliding bolides cause particle acceleration. In regions distinctly separate from the jet ridge line, the fractional polarisation approaches the theoretical maximum for synchrotron emission.

astro-ph.HE↗

arXiv2016

Anomaly Detection in Bitcoin Network Using Unsupervised Learning Methods

The problem of anomaly detection has been studied for a long time. In short, anomalies are abnormal or unlikely things. In financial networks, thieves and illegal activities are often anomalous in nature. Members of a network want to detect anomalies as soon as possible to prevent them from harming the network's community and integrity. Many Machine Learning techniques have been proposed to deal with this problem; some results appear to be quite promising but there is no obvious superior method. In this paper, we consider anomaly detection particular to the Bitcoin transaction network. Our goal is to detect which users and transactions are the most suspicious; in this case, anomalous behavior is a proxy for suspicious behavior. To this end, we use three unsupervised learning methods including k-means clustering, Mahalanobis distance, and Unsupervised Support Vector Machine (SVM) on two graphs generated by the Bitcoin transaction network: one graph has users as nodes, and the other has transactions as nodes.

cs.LG↗

arXiv2016

Anomaly Detection in the Bitcoin System - A Network Perspective

The problem of anomaly detection has been studied for a long time, and many Network Analysis techniques have been proposed as solutions. Although some results appear to be quite promising, no method is clearly to be superior to the rest. In this paper, we particularly consider anomaly detection in the Bitcoin transaction network. Our goal is to detect which users and transactions are the most suspicious; in this case, anomalous behavior is a proxy for suspicious behavior. To this end, we use the laws of power degree and densification and local outlier factor (LOF) method (which is proceeded by k-means clustering method) on two graphs generated by the Bitcoin transaction network: one graph has users as nodes, and the other has transactions as nodes. We remark that the methods used here can be applied to any type of setting with an inherent graph structure, including, but not limited to, computer networks, telecommunications networks, auction networks, security networks, social networks, Web networks, or any financial networks. We use the Bitcoin transaction network in this paper due to the availability, size, and attractiveness of the data set.

cs.SI↗

arXiv2009

An investigation of lucky imaging techniques

We present an empirical analysis of the effectiveness of frame selection (also known as Lucky Imaging) techniques for high resolution imaging. A high-speed image recording system has been used to observe a number of bright stars. The observations were made over a wide range of values of D/r0 and exposure time. The improvement in Strehl ratio of the stellar images due to aligning frames and selecting the best frames was evaluated as a function of these parameters. We find that improvement in Strehl ratio by factors of 4 to 6 can be achieved over a range of D/r0 from 3 to 12, with a slight peak at D/r0 ~ 7. The best Strehl improvement is achieved with exposure times of 10 ms or less but significant improvement is still obtained at exposure times as long as 640 ms. Our results are consistent with previous investigations but cover a much wider range of parameter space. We show that Strehl ratios of >0.7 can be achieved in appropiate conditions whereas previous studies have generally shown maximum Strehl ratios of ~0.3. The results are in reasonable agreement with the simulations of Baldwin et al. (2008).

astro-ph.IM↗