Abstract
Label noise is omnipresent in the annotations process and has an impact on supervised learning algorithms. This work focuses on the impact of label noise on the performance of learning models by examining the effect of random and class-dependent label noise on a binary classification task: quality assessment for photoplethysmography (PPG). PPG signal is used to detect physiological changes and its quality can have a significant impact on the subsequent tasks, which makes PPG quality assessment a particularly good target for examining the impact of label noise in the field of biomedicine. Random and class-dependent label noise was introduced separately into the training set to emulate the errors associated with fatigue and bias in labeling data samples. We also tested different representations of the PPG, including features defined by domain experts, 1D raw signal and 2D image. Three different classifiers are tested on the noisy training data, including support vector machine (SVM), XGBoost, 1D Resnet and 2D Resnet, which handle three representations, respectively. The results showed that the two deep learning models were more robust than the two traditional machine learning models for both the random and class-dependent label noise. From the representation perspective, the 2D image shows better robustness compared to the 1D raw signal. The logits from three classifiers are also analyzed, the predicted probabilities intend to be more dispersed when more label noise is introduced. From this work, we investigated various factors related to label noise, including representations, label noise type, and data imbalance, which can be a good guidebook for designing more robust methods for label noise in future work.

Abstract

Abstract
Context. The low wind effect (LWE) occurs at the aperture of 8-meter class telescopes when the spiders holding the secondary mirror get significantly cooler than the air. The effect creates phase discontinuities in the incoming wavefront at the location of the spiders. Under the LWE, the wavefront residuals after correction of the adaptive optics (AO) are dominated by low-order aberrations, pistons, and tip-tilts, contained in the pupil quadrants separated by the spiders. Those aberrations, called petal modes, degrade the AO performances during the best atmospheric turbulence conditions. Ultimately, the LWE is an obstacle for high-contrast exoplanet observations at a small angular separation from the host star. Aims. We aim to understand why extreme AO with a Shack-Hartmann (SH) wavefront sensor fails to correct for the petal tip and tilt modes, while these modes imprint a measurable signal in the SH slopes. We explore if the petal tip and tilt content of the LWE can be controlled and mitigated without an additional wavefront sensor. Methods. We simulated the sensitivity of a single subaperture of a SH wavefront sensor in the presence of a phase discontinuity across this subaperture. We explored the effect of the most important parameters: the amplitude of the discontinuity, the spider thickness, and the field of view. We then performed end-to-end simulations to reproduce and explain the behavior of extreme AO systems based on a SH in the presence of the LWE. We then evaluated the efficiency of a new mitigation strategy by running simulations, including atmosphere and realistic LWE phase perturbations. Results. For realistic parameters (i.e. a spider thickness at 25% of a SH subaperture, and a field of view of 3.5 lambda/d), we find that the sensitivity of the SH to a phase discontinuity is dramatically reduced, or even reversed. Under the LWE, a nonzero curl path is created in the measured slopes, which transforms into vortex-structures in the residuals when the loop is closed. While these vortexes are easily seen in the residual wavefront and slopes, they cannot be controlled by the system. We used this understanding to propose a strategy for controlling the petal tip and tilt modes of the LWE by using the measurements from the SH, but excluding the faulty subapertures. Conclusions. The proposed mitigation strategy may be of use in all extreme AO systems based on SH for which the LWE is an issue, such as SPHERE and GRAVITY+.

Abstract
The W. M. Keck Observatory Adaptive Optics (AO) facilities have been operating with a Field Programmable Gate Array (FPGA) based real time controller (RTC) since 2007. The RTC inputs data from various AO wavefront and tip/tilt sensors; and corrects image blurring from atmospheric turbulence via deformable and tip/tilt mirrors. Since its commissioning, the Keck I and Keck II RTCs have been upgraded to support new hardware such as pyramid wavefront and infrared tip-tilt sensors. However, they are reaching the limits of their capabilities in terms of processing bandwidth and the ability to interface with new hardware. Together with the Keck All-sky Precision Adaptive optics (KAPA) project, a higher performance and a more reliable RTC is needed to support next generation capabilities such as laser tomography and sensor fusion. This paper provides an overview of the new RTC system, developed with our contractor/collaborators (Microgate, Swinburne University of Technology and Australian National University), and the initial on-sky performance. The upgrade includes an Interface Module to interface with the wavefront sensors and controlled hardware, and a Graphical Processing Unit (GPU) based computational engine to meet the system's control requirements and to provide a flexible software architecture to allow future algorithms development and capabilities. The system saw first light in 2021 and is being commissioned in 2022 to support single conjugate laser guide star (LGS) AO, along with a more sensitive EMCCD camera. Initial results are provided to demonstrate single NGS & LGS performance, system reliability, and the planned upgrade for four LGS to support laser tomography.

Abstract
The Gemini Infra-Red Multi-Object Spectrograph (GIRMOS) is a four-arm, Multi-Object Adaptive Optics (MOAO) IFU spectrograph being built for Gemini (commissioning in 2024). GIRMOS is being planned to interface with the new Gemini-North Adaptive Optics (GNAO) system, and is base lined with a requirement of 50% EE within a 0.100 spaxel at H-band. We present a design and forecast the error budget and performance of GIRMOS-MOAO working behind GNAO. The MOAO system will patrol the 20 field of regard of GNAO, utilizing closed loop GLAO or MCAO for lower order correction. GIRMOS MOAA will perform tomographic reconstruction of the turbulence using the GNAO WFS, and utilize order 16x16 actuator DMs operating in open loop to perform an additional correction from the Pseudo Open Loop (POL) slopes, achieving close to diffraction limited performance from the combined GNAO+MOAO correction. This high performance AO spectrograph will have the broadest impact in the study of the formation and evolution of galaxies, but will also have broad reach in fields such as star and planet formation within our Milky Way and supermassive black holes in nearby galaxies.

Abstract
We present the status and plans for the Keck All sky Precision Adaptive optics (KAPA) program. KAPA includes (1) an upgrade to the Keck I laser guide star adaptive optics (AO) facility to improve image quality and sky coverage, (2) the inclusion of AO telemetry-based point spread function estimates with all science exposures, (3) four key science programs, and (4) an educational component focused on broadening the participation of women and underrepresented groups in instrumentation. For this conference we focus on the KAPA upgrades since the 2020 SPIE proceedings1 including implementation of a laser asterism generator, wavefront sensor, real-time controller, asterism and turbulence simulators, the laser tomography system itself along with new operations software and science tools, and modifications to an existing near-infrared tip-tilt sensor to support multiple natural guide star and focus measurements. We will also report on the results of daytime and on-sky calibrations and testing.