-
On the Out-of-Distribution Coverage of Combining Split Conformal Prediction and Bayesian Deep Learning
Authors:
Paul Scemama,
Ariel Kapusta
Abstract:
Bayesian deep learning and conformal prediction are two methods that have been used to convey uncertainty and increase safety in machine learning systems. We focus on combining Bayesian deep learning with split conformal prediction and how this combination effects out-of-distribution coverage; particularly in the case of multiclass image classification. We suggest that if the model is generally un…
▽ More
Bayesian deep learning and conformal prediction are two methods that have been used to convey uncertainty and increase safety in machine learning systems. We focus on combining Bayesian deep learning with split conformal prediction and how this combination effects out-of-distribution coverage; particularly in the case of multiclass image classification. We suggest that if the model is generally underconfident on the calibration set, then the resultant conformal sets may exhibit worse out-of-distribution coverage compared to simple predictive credible sets. Conversely, if the model is overconfident on the calibration set, the use of conformal prediction may improve out-of-distribution coverage. We evaluate prediction sets as a result of combining split conformal methods and neural networks trained with (i) stochastic gradient descent, (ii) deep ensembles, and (iii) mean-field variational inference. Our results suggest that combining Bayesian deep learning models with split conformal prediction can, in some cases, cause unintended consequences such as reducing out-of-distribution coverage.
△ Less
Submitted 7 March, 2024; v1 submitted 21 November, 2023;
originally announced November 2023.
-
Towards Reliable Colorectal Cancer Polyps Classification via Vision Based Tactile Sensing and Confidence-Calibrated Neural Networks
Authors:
Siddhartha Kapuria,
Tarunraj G. Mohanraj,
Nethra Venkatayogi,
Ozdemir Can Kara,
Yuki Hirata,
Patrick Minot,
Ariel Kapusta,
Naruhiko Ikoma,
Farshid Alambeigi
Abstract:
In this study, toward addressing the over-confident outputs of existing artificial intelligence-based colorectal cancer (CRC) polyp classification techniques, we propose a confidence-calibrated residual neural network. Utilizing a novel vision-based tactile sensing (VS-TS) system and unique CRC polyp phantoms, we demonstrate that traditional metrics such as accuracy and precision are not sufficien…
▽ More
In this study, toward addressing the over-confident outputs of existing artificial intelligence-based colorectal cancer (CRC) polyp classification techniques, we propose a confidence-calibrated residual neural network. Utilizing a novel vision-based tactile sensing (VS-TS) system and unique CRC polyp phantoms, we demonstrate that traditional metrics such as accuracy and precision are not sufficient to encapsulate model performance for handling a sensitive CRC polyp diagnosis. To this end, we develop a residual neural network classifier and address its over-confident outputs for CRC polyps classification via the post-processing method of temperature scaling. To evaluate the proposed method, we introduce noise and blur to the obtained textural images of the VS-TS and test the model's reliability for non-ideal inputs through reliability diagrams and other statistical metrics.
△ Less
Submitted 25 April, 2023;
originally announced April 2023.
-
Bodies at Rest: 3D Human Pose and Shape Estimation from a Pressure Image using Synthetic Data
Authors:
Henry M. Clever,
Zackory Erickson,
Ariel Kapusta,
Greg Turk,
C. Karen Liu,
Charles C. Kemp
Abstract:
People spend a substantial part of their lives at rest in bed. 3D human pose and shape estimation for this activity would have numerous beneficial applications, yet line-of-sight perception is complicated by occlusion from bedding. Pressure sensing mats are a promising alternative, but training data is challenging to collect at scale. We describe a physics-based method that simulates human bodies…
▽ More
People spend a substantial part of their lives at rest in bed. 3D human pose and shape estimation for this activity would have numerous beneficial applications, yet line-of-sight perception is complicated by occlusion from bedding. Pressure sensing mats are a promising alternative, but training data is challenging to collect at scale. We describe a physics-based method that simulates human bodies at rest in a bed with a pressure sensing mat, and present PressurePose, a synthetic dataset with 206K pressure images with 3D human poses and shapes. We also present PressureNet, a deep learning model that estimates human pose and shape given a pressure image and gender. PressureNet incorporates a pressure map reconstruction (PMR) network that models pressure image generation to promote consistency between estimated 3D body models and pressure image input. In our evaluations, PressureNet performed well with real data from participants in diverse poses, even though it had only been trained with synthetic data. When we ablated the PMR network, performance dropped substantially.
△ Less
Submitted 2 April, 2020;
originally announced April 2020.
-
Assistive Gym: A Physics Simulation Framework for Assistive Robotics
Authors:
Zackory Erickson,
Vamsee Gangaram,
Ariel Kapusta,
C. Karen Liu,
Charles C. Kemp
Abstract:
Autonomous robots have the potential to serve as versatile caregivers that improve quality of life for millions of people worldwide. Yet, conducting research in this area presents numerous challenges, including the risks of physical interaction between people and robots. Physics simulations have been used to optimize and train robots for physical assistance, but have typically focused on a single…
▽ More
Autonomous robots have the potential to serve as versatile caregivers that improve quality of life for millions of people worldwide. Yet, conducting research in this area presents numerous challenges, including the risks of physical interaction between people and robots. Physics simulations have been used to optimize and train robots for physical assistance, but have typically focused on a single task. In this paper, we present Assistive Gym, an open source physics simulation framework for assistive robots that models multiple tasks. It includes six simulated environments in which a robotic manipulator can attempt to assist a person with activities of daily living (ADLs): itch scratching, drinking, feeding, body manipulation, dressing, and bathing. Assistive Gym models a person's physical capabilities and preferences for assistance, which are used to provide a reward function. We present baseline policies trained using reinforcement learning for four different commercial robots in the six environments. We demonstrate that modeling human motion results in better assistance and we compare the performance of different robots. Overall, we show that Assistive Gym is a promising tool for assistive robotics research.
△ Less
Submitted 10 October, 2019;
originally announced October 2019.
-
3D Human Pose Estimation on a Configurable Bed from a Pressure Image
Authors:
Henry M. Clever,
Ariel Kapusta,
Daehyung Park,
Zackory Erickson,
Yash Chitalia,
Charles C. Kemp
Abstract:
Robots have the potential to assist people in bed, such as in healthcare settings, yet bedding materials like sheets and blankets can make observation of the human body difficult for robots. A pressure-sensing mat on a bed can provide pressure images that are relatively insensitive to bedding materials. However, prior work on estimating human pose from pressure images has been restricted to 2D pos…
▽ More
Robots have the potential to assist people in bed, such as in healthcare settings, yet bedding materials like sheets and blankets can make observation of the human body difficult for robots. A pressure-sensing mat on a bed can provide pressure images that are relatively insensitive to bedding materials. However, prior work on estimating human pose from pressure images has been restricted to 2D pose estimates and flat beds. In this work, we present two convolutional neural networks to estimate the 3D joint positions of a person in a configurable bed from a single pressure image. The first network directly outputs 3D joint positions, while the second outputs a kinematic model that includes estimated joint angles and limb lengths. We evaluated our networks on data from 17 human participants with two bed configurations: supine and seated. Our networks achieved a mean joint position error of 77 mm when tested with data from people outside the training set, outperforming several baselines. We also present a simple mechanical model that provides insight into ambiguity associated with limbs raised off of the pressure mat, and demonstrate that Monte Carlo dropout can be used to estimate pose confidence in these situations. Finally, we provide a demonstration in which a mobile manipulator uses our network's estimated kinematic model to reach a location on a person's body in spite of the person being seated in a bed and covered by a blanket.
△ Less
Submitted 29 August, 2018; v1 submitted 20 April, 2018;
originally announced April 2018.
-
Task-centric Optimization of Configurations for Assistive Robots
Authors:
Ariel Kapusta,
Charles C. Kemp
Abstract:
Robots can provide assistance to a human by moving objects to locations around the person's body. With a well chosen initial configuration, a robot can better reach locations important to an assistive task despite model error, pose uncertainty and other sources of variation. However, finding effective configurations can be challenging due to complex geometry, a large number of degrees of freedom,…
▽ More
Robots can provide assistance to a human by moving objects to locations around the person's body. With a well chosen initial configuration, a robot can better reach locations important to an assistive task despite model error, pose uncertainty and other sources of variation. However, finding effective configurations can be challenging due to complex geometry, a large number of degrees of freedom, task complexity and other factors. We present task-centric optimization of robot configurations (TOC), which is an algorithm that finds configurations from which the robot can better reach task-relevant locations and handle task variation. Notably, TOC can return more than one configuration that when used sequentially enable a simulated assistive robot to reach more task-relevant locations. TOC performs substantial offline computation to generate a function that can be applied rapidly online to select robot configurations based on current observations. TOC explicitly models the task, environment, and user, and implicitly handles error using representations of robot dexterity. We evaluated TOC in simulation with a PR2 assisting a user with 9 assistive tasks in both a wheelchair and a robotic bed. TOC had an overall average success rate of 90.6\% compared to 50.4\%, 43.5\%, and 58.9\% for three baseline methods from literature. We additionally demonstrate how TOC can find configurations for more than one robot and can be used to assist in designing or optimizing environments.
△ Less
Submitted 19 April, 2018;
originally announced April 2018.
-
Tracking Human Pose During Robot-Assisted Dressing using Single-Axis Capacitive Proximity Sensing
Authors:
Zackory Erickson,
Maggie Collier,
Ariel Kapusta,
Charles C. Kemp
Abstract:
Dressing is a fundamental task of everyday living and robots offer an opportunity to assist people with motor impairments. While several robotic systems have explored robot-assisted dressing, few have considered how a robot can manage errors in human pose estimation, or adapt to human motion in real time during dressing assistance. In addition, estimating pose changes due to human motion can be ch…
▽ More
Dressing is a fundamental task of everyday living and robots offer an opportunity to assist people with motor impairments. While several robotic systems have explored robot-assisted dressing, few have considered how a robot can manage errors in human pose estimation, or adapt to human motion in real time during dressing assistance. In addition, estimating pose changes due to human motion can be challenging with vision-based techniques since dressing is often intended to visually occlude the body with clothing. We present a method to track a person's pose in real time using capacitive proximity sensing. This sensing approach gives direct estimates of distance with low latency, has a high signal-to-noise ratio, and has low computational requirements. Using our method, a robot can adjust for errors in the estimated pose of a person and physically follow the contours and movements of the person while providing dressing assistance. As part of an evaluation of our method, the robot successfully pulled the sleeve of a hospital gown and a cardigan onto the right arms of 10 human participants, despite arm motions and large errors in the initially estimated pose of the person's arm. We also show that a capacitive sensor is unaffected by visual occlusion of the body and can sense a person's body through cotton clothing.
△ Less
Submitted 24 May, 2019; v1 submitted 22 September, 2017;
originally announced September 2017.
-
Lineage-specific transposons drove massive gene expression recruitments during the evolution of pregnancy in mammals
Authors:
Vincent J. Lynch,
Mauris Nnamani,
Kathryn J. Brayer,
Deena Emera,
Joel O. Wertheim,
Sergei L. Kosakovsky Pond,
Frank Grützner,
Stefan Bauersachs,
Alexander Graf,
Aurélie Kapusta,
Cédric Feschotte,
Günter P. Wagner
Abstract:
A major challenge in biology is explaining how novel characters originate, however, the molecular mechanisms that underlie the emergence of evolutionary innovations are unclear. Here we show that while gene expression in the uterus evolves at a slow and relatively constant rate, it has been punctuated by periods of rapid change associated with the recruitment of thousands of genes into uterine exp…
▽ More
A major challenge in biology is explaining how novel characters originate, however, the molecular mechanisms that underlie the emergence of evolutionary innovations are unclear. Here we show that while gene expression in the uterus evolves at a slow and relatively constant rate, it has been punctuated by periods of rapid change associated with the recruitment of thousands of genes into uterine expression during the evolution of pregnancy in mammals. We found that numerous genes and signaling pathways essential for the establishment of pregnancy and maternal-fetal communication evolved uterine expression in mammals. Remarkably the majority of genes recruited into endometrial expression have cis-regulatory elements derived from lineage-specific transposons, suggesting that that bursts of transposition facilitate adaptation and speciation through genomic and regulatory reorganization.
△ Less
Submitted 22 August, 2012;
originally announced August 2012.
-
Spectroscopy of Nine Cataclysmic Variable Stars
Authors:
H. A. Sheets,
J. R. Thorstensen,
C. J. Peters,
A. B. Kapusta,
C. J. Taylor
Abstract:
We present optical spectroscopy of nine cataclysmic binary stars, mostly dwarf novae, obtained primarily to determine orbital periods Porb. The stars and their periods are LX And, 0.1509743(5) d; CZ Aql, 0.2005(6) d; LU Cam, 0.1499686(4) d; GZ Cnc, 0.0881(4) d; V632 Cyg, 0.06377(8) d; V1006 Cyg, 0.09903(9) d; BF Eri, 0.2708804(4) d; BI Ori, 0.1915(5) d; and FO Per, for which Porb is either 0.146…
▽ More
We present optical spectroscopy of nine cataclysmic binary stars, mostly dwarf novae, obtained primarily to determine orbital periods Porb. The stars and their periods are LX And, 0.1509743(5) d; CZ Aql, 0.2005(6) d; LU Cam, 0.1499686(4) d; GZ Cnc, 0.0881(4) d; V632 Cyg, 0.06377(8) d; V1006 Cyg, 0.09903(9) d; BF Eri, 0.2708804(4) d; BI Ori, 0.1915(5) d; and FO Per, for which Porb is either 0.1467(4) or 0.1719(5) d.
Several of the stars proved to be especially interesting. In BF Eri, we detect the absorption spectrum of a secondary star of spectral type K3 +- 1 subclass, which leads to a distance estimate of approximately 1 kpc. However, BF Eri has a large proper motion (100 mas/yr), and we have a preliminary parallax measurement that confirms the large proper motion and yields only an upper limit for the parallax. BF Eri's space velocity is evidently large, and it appears to belong to the halo population. In CZ Aql, the emission lines have strong wings that move with large velocity amplitude, suggesting a magnetically-channeled accretion flow. The orbital period of V1006 Cyg places it squarely within the 2- to 3-hour "gap" in the distribution of cataclysmic binary orbital periods.
△ Less
Submitted 6 April, 2007;
originally announced April 2007.
-
Orbital Period of the Dwarf Nova RXS J053234.9+624755
Authors:
Ann B. Kapusta,
John R. Thorstensen
Abstract:
We report spectroscopy of the newly discovered SU Ursae Majoris dwarf nova identified with the x-ray source RXS J053234.9+624755. Radial velocities of the H-alpha emission line in the quiescent state give an orbital period of 0.05620(4) d (80.93 min), which is among the shortest for SU UMa stars with determined periods. We also report UBVI magnitudes of the quiescent dwarf nova and surrounding s…
▽ More
We report spectroscopy of the newly discovered SU Ursae Majoris dwarf nova identified with the x-ray source RXS J053234.9+624755. Radial velocities of the H-alpha emission line in the quiescent state give an orbital period of 0.05620(4) d (80.93 min), which is among the shortest for SU UMa stars with determined periods. We also report UBVI magnitudes of the quiescent dwarf nova and surrounding stars. Using a previous measurement of the superhump period, we find the fractional superhump excess epsilon to be 0.016(4), which is not atypical of dwarf novae in this period range.
△ Less
Submitted 5 July, 2006;
originally announced July 2006.
