-
Bilevel methods for image reconstruction
Authors:
Caroline Crockett,
Jeffrey A. Fessler
Abstract:
This review discusses methods for learning parameters for image reconstruction problems using bilevel formulations. Image reconstruction typically involves optimizing a cost function to recover a vector of unknown variables that agrees with collected measurements and prior assumptions. State-of-the-art image reconstruction methods learn these prior assumptions from training data using various mach…
▽ More
This review discusses methods for learning parameters for image reconstruction problems using bilevel formulations. Image reconstruction typically involves optimizing a cost function to recover a vector of unknown variables that agrees with collected measurements and prior assumptions. State-of-the-art image reconstruction methods learn these prior assumptions from training data using various machine learning techniques, such as bilevel methods. One can view the bilevel problem as formalizing hyperparameter optimization, as bridging machine learning and cost function based optimization methods, or as a method to learn variables best suited to a specific task. More formally, bilevel problems attempt to minimize an upper-level loss function, where variables in the upper-level loss function are themselves minimizers of a lower-level cost function.
This review contains a running example problem of learning tuning parameters and the coefficients for sparsifying filters used in a regularizer. Such filters generalize the popular total variation regularization method, and learned filters are closely related to convolutional neural networks approaches that are rapidly gaining in popularity. Here, the lower-level problem is to reconstruct an image using a regularizer with learned sparsifying filters; the corresponding upper-level optimization problem involves a measure of reconstructed image quality based on training data. This review discusses multiple perspectives to motivate the use of bilevel methods and to make them more easily accessible to different audiences. We then turn to ways to optimize the bilevel problem, providing pros and cons of the variety of proposed approaches. Finally we overview bilevel applications in image reconstruction.
△ Less
Submitted 15 June, 2022; v1 submitted 20 September, 2021;
originally announced September 2021.
-
A tidal lung simulation to quantify lung heterogeneity with the Inspired Sinewave Test
Authors:
Minh C. Tran,
Douglas C. Crockett,
Phi A. Phan,
Stephen J. Payne,
Andrew D. Farmery
Abstract:
We have created a lung simulation to quantify lung heterogeneity from the results of the inspired sinewave test (IST). The IST is a lung function test that is non-invasive, non-ionising and does not require patients' cooperation. A tidal lung simulation is developed to assess this test and also a method is proposed to calculate lung heterogeneity from IST results. A sensitivity analysis based on t…
▽ More
We have created a lung simulation to quantify lung heterogeneity from the results of the inspired sinewave test (IST). The IST is a lung function test that is non-invasive, non-ionising and does not require patients' cooperation. A tidal lung simulation is developed to assess this test and also a method is proposed to calculate lung heterogeneity from IST results. A sensitivity analysis based on the Morris method and linear regression were applied to verify and to validate the simulation. Additionally, simulated emphysema and pulmonary embolism conditions were created using the simulation to assess the ability of the IST to identify these conditions. Experimental data from five pigs (pre-injured vs injured) were used for validation. This paper contributes to the development of the IST. Firstly, our sensitivity analysis reveals that the IST is highly accurate with an underestimation of about 5% of the simulated values. Sensitivity analysis suggested that both instability in tidal volume and extreme expiratory flow coefficients during the test cause random errors in the IST results. Secondly, the ratios of IST results obtained at two tracer gas oscillation frequencies can identify lung heterogeneity (ELV60/ELV180 and Qp60/Qp180). There was dissimilarity between simulated emphysema and pulmonary embolism (p < 0:0001). In the animal model, the control group had ELV60/ELV180 = 0.58 compared with 0.39 in injured animals (p < 0.0001).
△ Less
Submitted 29 September, 2020;
originally announced September 2020.
-
A Candidate Young Massive Planet in Orbit around the Classical T Tauri Star CI Tau
Authors:
Christopher M. Johns-Krull,
Jacob N. McLane,
L. Prato,
Christopher J. Crockett,
Daniel T. Jaffe,
Patrick M. Hartigan,
Charles A. Beichman,
Naved I. Mahmud,
Wei Chen,
B. A. Skiff,
P. Wilson Cauley,
Joshua A. Jones,
G. N. Mace
Abstract:
The ~2 Myr old classical T Tauri star CI Tau shows periodic variability in its radial velocity (RV) variations measured at infrared (IR) and optical wavelengths. We find that these observations are consistent with a massive planet in a ~9-day period orbit. These results are based on 71 IR RV measurements of this system obtained over 5 years, and on 26 optical RV measurements obtained over 9 years.…
▽ More
The ~2 Myr old classical T Tauri star CI Tau shows periodic variability in its radial velocity (RV) variations measured at infrared (IR) and optical wavelengths. We find that these observations are consistent with a massive planet in a ~9-day period orbit. These results are based on 71 IR RV measurements of this system obtained over 5 years, and on 26 optical RV measurements obtained over 9 years. CI Tau was also observed photometrically in the optical on 34 nights over ~one month in 2012. The optical RV data alone are inadequate to identify an orbital period, likely the result of star spot and activity induced noise for this relatively small dataset. The infrared RV measurements reveal significant periodicity at ~9 days. In addition, the full set of optical and IR RV measurements taken together phase coherently and with equal amplitudes to the ~9 day period. Periodic radial velocity signals can in principle be produced by cool spots, hot spots, and reflection of the stellar spectrum off the inner disk, in addition to resulting from a planetary companion. We have considered each of these and find the planet hypothesis most consistent with the data. The radial velocity amplitude yields an Msin(i) of ~8.1 M_Jup; in conjunction with a 1.3 mm continuum emission measurement of the circumstellar disk inclination from the literature, we find a planet mass of ~11.3 M_Jup, assuming alignment of the planetary orbit with the disk.
△ Less
Submitted 25 May, 2016;
originally announced May 2016.
-
The First U.S. Naval Observatory Robotic Astrometric Telescope Catalog (URAT1)
Authors:
Norbert Zacharias,
Charlie Finch,
John Subasavage,
Greg Bredthauer,
Chris Crockett,
Mike DiVittorio,
Erik Ferguson,
Fred Harris,
Hugh Harris,
Arne Henden,
Chris Kilian,
Jeff Munn,
Ted Rafferty,
Al Rhodes,
Mike Schultheiss,
Trudy Tilleman,
Gary Wieder
Abstract:
URAT1 is an observational, astrometric catalog covering most of the Dec >= -15 deg area and a magnitude range of about R = 3 to 18.5. Accurate positions (typically 10 to 30 mas standard error) are given for over 228 million objects at a mean epoch around 2013.5. For the over 188 million objects matched with the 2MASS point source catalog proper motions (typically 5 to 7 mas/yr std. errors) are pro…
▽ More
URAT1 is an observational, astrometric catalog covering most of the Dec >= -15 deg area and a magnitude range of about R = 3 to 18.5. Accurate positions (typically 10 to 30 mas standard error) are given for over 228 million objects at a mean epoch around 2013.5. For the over 188 million objects matched with the 2MASS point source catalog proper motions (typically 5 to 7 mas/yr std. errors) are provided. These data are supplemented by 2MASS and APASS photometry. Observations, reductions and catalog construction are described together with results from external data verifications. The catalog data are served by CDS, Starsbourg (I/329). There is no DVD release.
△ Less
Submitted 19 August, 2015;
originally announced August 2015.
-
Follow-Up Observations of PTFO 8-8695: A 3 MYr Old T-Tauri Star Hosting a Jupiter-mass Planetary Candidate
Authors:
David R. Ciardi,
J. C. van Eyken,
J. W. Barnes,
C. A. Beichman,
S. J. Carey,
C. J. Crockett,
J. Eastman,
C. M. Johns-Krull,
S. B. Howell,
S. R. Kane,
J. N. Mclane,
P. Plavchan,
L. Prato,
J. Stauffer,
G. T. van Belle,
K. von Braun
Abstract:
We present Spitzer 4.5\micron\ light curve observations, Keck NIRSPEC radial velocity observations, and LCOGT optical light curve observations of PTFO~8-8695, which may host a Jupiter-sized planet in a very short orbital period (0.45 days). Previous work by \citet{vaneyken12} and \citet{barnes13} predicts that the stellar rotation axis and the planetary orbital plane should precess with a period o…
▽ More
We present Spitzer 4.5\micron\ light curve observations, Keck NIRSPEC radial velocity observations, and LCOGT optical light curve observations of PTFO~8-8695, which may host a Jupiter-sized planet in a very short orbital period (0.45 days). Previous work by \citet{vaneyken12} and \citet{barnes13} predicts that the stellar rotation axis and the planetary orbital plane should precess with a period of $300 - 600$ days. As a consequence, the observed transits should change shape and depth, disappear, and reappear with the precession. Our observations indicate the long-term presence of the transit events ($>3$ years), and that the transits indeed do change depth, disappear and reappear. The Spitzer observations and the NIRSPEC radial velocity observations (with contemporaneous LCOGT optical light curve data) are consistent with the predicted transit times and depths for the $M_\star = 0.34\ M_\odot$ precession model and demonstrate the disappearance of the transits. An LCOGT optical light curve shows that the transits do reappear approximately 1 year later. The observed transits occur at the times predicted by a straight-forward propagation of the transit ephemeris. The precession model correctly predicts the depth and time of the Spitzer transit and the lack of a transit at the time of the NIRSPEC radial velocity observations. However, the precession model predicts the return of the transits approximately 1 month later than observed by LCOGT. Overall, the data are suggestive that the planetary interpretation of the observed transit events may indeed be correct, but the precession model and data are currently insufficient to confirm firmly the planetary status of PTFO~8-8695b.
△ Less
Submitted 29 June, 2015;
originally announced June 2015.
-
The Closest M-Dwarf Quadruple System to the Sun
Authors:
Cassy L. Davison,
Russel J. White,
Wei-Chun Jao,
Todd J. Henry,
John I. Bailey III,
Samuel N. Quinn,
Justin R. Cantrell,
Adric R. Riedel,
John P. Subasavage,
Jen G. Winters,
Christopher J. Crockett
Abstract:
We report new infrared radial velocity measurements obtained with CSHELL at NASA's Infrared Telescope Facility that reveal the M3.5 dwarf GJ 867B to be a single-lined spectroscopic binary with a period of 1.795 $\pm$ 0.017 days. Its velocity semi-amplitude of 21.4 $\pm$ 0.5 km s$^{-1}$ corresponds to a minimum mass of 61 $\pm$ 7 M$_{JUP}$; the new companion, which we call GJ 867D, could be a brown…
▽ More
We report new infrared radial velocity measurements obtained with CSHELL at NASA's Infrared Telescope Facility that reveal the M3.5 dwarf GJ 867B to be a single-lined spectroscopic binary with a period of 1.795 $\pm$ 0.017 days. Its velocity semi-amplitude of 21.4 $\pm$ 0.5 km s$^{-1}$ corresponds to a minimum mass of 61 $\pm$ 7 M$_{JUP}$; the new companion, which we call GJ 867D, could be a brown dwarf. Stable astrometric measurements of GJ 867BD obtained with CTIO's 0.9-m telescope over the last decade exclude the presence of any massive planetary companions (7-18 M$_{JUP}$) with longer orbital periods (2-10 years) for the majority of orientations. These complementary observations are also used to determine the trigonometric distance and proper motion of GJ 867BD; the measurements are consistent with the HIPPARCOS measured values of the M2 dwarf GJ 867AC, which is itself a 4.1 day double-lined spectroscopic binary at a projected separation of 24.5" (216 AU) from GJ 867BD. These new measurements strengthen the case that GJ 867AC and GJ 867BD are physically associated, making the GJ 867 system one of only four quadruple systems within 10 pc of the Sun (d$=$ 8.82 $\pm$0.08 pc) and the only among these with all M-dwarf (or cooler) components.
△ Less
Submitted 24 October, 2013;
originally announced October 2013.
-
A Search for Giant Planet Companions to T Tauri Stars
Authors:
Christopher J. Crockett,
Naved I. Mahmud,
L. Prato,
Christopher M. Johns-Krull,
Daniel T. Jaffe,
Patrick M. Hartigan,
Charles A. Beichman
Abstract:
We present results from an ongoing multiwavelength radial velocity (RV) survey of the Taurus-Auriga star forming region as part of our effort to identify pre--main sequence giant planet hosts. These 1-3 Myr old T Tauri stars present significant challenges to traditional RV surveys. The presence of strong magnetic fields gives rise to large, cool star spots. These spots introduce significant RV jit…
▽ More
We present results from an ongoing multiwavelength radial velocity (RV) survey of the Taurus-Auriga star forming region as part of our effort to identify pre--main sequence giant planet hosts. These 1-3 Myr old T Tauri stars present significant challenges to traditional RV surveys. The presence of strong magnetic fields gives rise to large, cool star spots. These spots introduce significant RV jitter which can mimic the velocity modulation from a planet-mass companion. To distinguish between spot-induced and planet-induced RV modulation, we conduct observations at ~6700 Angstroms and ~2.3 microns and measure the wavelength dependence (if any) in the RV amplitude. CSHELL observations of the known exoplanet host Gl 86 demonstrate our ability to detect not only hot Jupiters in the near infrared but also secular trends from more distant companions. Observations of nine very young stars reveal a typical reduction in RV amplitude at the longer wavelengths by a factor of ~2-3. While we can not confirm the presence of planets in this sample, three targets show different periodicities in the two wavelength regions. This suggests different physical mechanisms underlying the optical and K band variability.
△ Less
Submitted 13 November, 2012; v1 submitted 6 November, 2012;
originally announced November 2012.
-
Starspot-induced optical and infrared radial velocity variability in T Tauri star Hubble 4
Authors:
Naved I. Mahmud,
Christopher J. Crockett,
Christopher M. Johns-Krull,
L. Prato,
Patrick M. Hartigan,
Daniel T. Jaffe,
Charles A. Beichman
Abstract:
We report optical (6150 Ang) and K-band (2.3 micron) radial velocities obtained over two years for the pre-main sequence weak-lined T Tauri star Hubble I 4. We detect periodic and near-sinusoidal radial velocity variations at both wavelengths, with a semi-amplitude of 1395\pm94 m/s in the optical and 365\pm80 m/s in the infrared. The lower velocity amplitude at the longer wavelength, combined with…
▽ More
We report optical (6150 Ang) and K-band (2.3 micron) radial velocities obtained over two years for the pre-main sequence weak-lined T Tauri star Hubble I 4. We detect periodic and near-sinusoidal radial velocity variations at both wavelengths, with a semi-amplitude of 1395\pm94 m/s in the optical and 365\pm80 m/s in the infrared. The lower velocity amplitude at the longer wavelength, combined with bisector analysis and spot modeling, indicates that there are large, cool spots on the stellar surface that are causing the radial velocity modulation. The radial velocities maintain phase coherence over hundreds of days suggesting that the starspots are long-lived. This is one of the first active stars where the spot-induced velocity modulation has been resolved in the infrared.
△ Less
Submitted 11 May, 2011;
originally announced May 2011.
-
Precision radial velocities with CSHELL
Authors:
Christopher J. Crockett,
Naved I. Mahmud,
Lisa Prato,
Christopher M. Johns-Krull,
Daniel T. Jaffe,
Charles A. Beichman
Abstract:
Radial velocity identification of extrasolar planets has historically been dominated by optical surveys. Interest in expanding exoplanet searches to M dwarfs and young stars, however, has motivated a push to improve the precision of near infrared radial velocity techniques. We present our methodology for achieving 58 m/s precision in the K band on the M0 dwarf GJ 281 using the CSHELL spectrograph…
▽ More
Radial velocity identification of extrasolar planets has historically been dominated by optical surveys. Interest in expanding exoplanet searches to M dwarfs and young stars, however, has motivated a push to improve the precision of near infrared radial velocity techniques. We present our methodology for achieving 58 m/s precision in the K band on the M0 dwarf GJ 281 using the CSHELL spectrograph at the 3-meter NASA IRTF. We also demonstrate our ability to recover the known 4 Mjup exoplanet Gl 86 b and discuss the implications for success in detecting planets around 1-3 Myr old T Tauri stars.
△ Less
Submitted 19 April, 2011;
originally announced April 2011.
-
Digital image exploration at Maui Community College
Authors:
Katie M. Morzinski,
Christopher J. Crockett,
Ian J. Crossfield
Abstract:
We designed a two-day laboratory exploration of fundamental concepts in digital images for an introductory engineering course at Maui Community College. Our objective was for the students to understand spatial vs. brightness resolution, standard file formats, image tradeoffs, and the engineering design cycle. We used open investigation, question generation, and an engineering design challenge to h…
▽ More
We designed a two-day laboratory exploration of fundamental concepts in digital images for an introductory engineering course at Maui Community College. Our objective was for the students to understand spatial vs. brightness resolution, standard file formats, image tradeoffs, and the engineering design cycle. We used open investigation, question generation, and an engineering design challenge to help our students achieve these learning goals. We also experimented with incorporating Hawaiian language and cultural awareness into our activity. We present our method, student response, and reflections on the success of our design. The 2008 re-design of this activity focused on better incorporating authentic engineering process skills, and on using a rubric for summative assessment of the students' poster presentations. A single file containing all documents and presentations used in this lesson is available online.
△ Less
Submitted 16 September, 2010;
originally announced September 2010.
-
The Formation and Evolution of Planetary Systems: The Search for and Characterization of Young Planets
Authors:
Charles Beichman,
Isabelle Baraffe,
Christopher Crockett,
Sarah Dodson-Robinson,
Jonathan Fortney,
Andrea Ghez,
Thomas P. Greene,
Adam Kraus,
Doug Lin,
Naved Mahmud,
Fabien Malbet,
Mark Marley,
Rafael Millan-Gabet,
Lisa Prato,
Ben Oppenheimer,
Michal Simon,
John Stauffer,
Angelle Tanner,
T. Velusamy
Abstract:
Despite the revolution in our knowledge resulting from the detection of planets around mature stars, we know almost nothing about planets orbiting young stars because rapid rotation and active photospheres preclude detection by radial velocities or transits and because direct imaging has barely penetrated the requisite range of high contrast and angular resolution. Of the techniques presently un…
▽ More
Despite the revolution in our knowledge resulting from the detection of planets around mature stars, we know almost nothing about planets orbiting young stars because rapid rotation and active photospheres preclude detection by radial velocities or transits and because direct imaging has barely penetrated the requisite range of high contrast and angular resolution. Of the techniques presently under consideration for the coming decade, only space-based astrometry offers the prospect of discovering gas giants (100 to >> 300 Mearth), lower mass systems such as icy giants (10 to 100 Mearth), and even a few rocky, super-Earths 300 Mearth) orbiting stars ranging in age from 1 to 100 Myr. Astrometry will complement high contrast imaging which should be able to detect gas giants (1~10 MJup) in orbits from a few to a few hundred AU. An astrometric survey in combination with imaging data for a subsample of objects will allow a detailed physical understanding of the formation and evolution of young gas giant planets impossible to achieve by any one technique.
△ Less
Submitted 17 February, 2009;
originally announced February 2009.
