Papers made digestable
Our architecture simplifies the obstacle-perception
problem to that of place-dependent change detection. While we use the method with VT&R, it
can be generalized to suit arbitrary path-following applications.
Visual Teach and Repeat 3 (VT&R3), a generalization of stereo VT&R, achieves
long-term autonomous path-following using topometric mapping and localization
from a single rich sensor stream. In this paper, we improve the capabilities of
a LiDAR implementation of VT&R3 to reliably detect and avoid obstacles in
changing environments. Our architecture simplifies the obstacle-perception
problem to that of place-dependent change detection. We then extend the
behaviour of generic sample-based motion planners to better suit the
teach-and-repeat problem structure by introducing a new edge-cost metric paired
with a curvilinear planning space. The resulting planner generates naturally
smooth paths that avoid local obstacles while minimizing lateral path deviation
to best exploit prior terrain knowledge. While we use the method with VT&R, it
can be generalized to suit arbitrary path-following applications. Experimental
results from online run-time analysis, unit testing, and qualitative
experiments on a differential drive robot show the promise of the technique for
reliable long-term autonomous operation in complex unstructured environments.
Authors: Jordy Sehn, Yuchen Wu, Timothy D. Barfoot.
The statistical and design considerations that pertain to
dose optimization are discussed. The sample size savings range from 16.6% to 27.3%,
depending on the design and scenario, with a mean savings of 22.1%.
The traditional more-is-better dose selection paradigm, developed based on
cytotoxic chemotherapeutics, is often problematic When applied to the
development of novel molecularly targeted agents (e.g., kinase inhibitors,
monoclonal antibodies, and antibody-drug conjugates). The US Food and Drug
Administration (FDA) initiated Project Optimus to reform the dose optimization
and dose selection paradigm in oncology drug development and call for more
attention to benefit-risk consideration.
We systematically investigated the operating characteristics of the seamless
phase 2-3 design as a strategy for dose optimization, where in stage 1
(corresponding to phase 2) patients are randomized to multiple doses, with or
without a control; and in stage 2 (corresponding to phase 3) the efficacy of
the selected optimal dose is evaluated with a randomized concurrent control or
historical control. Depending on whether the concurrent control is included and
the type of endpoints used in stages 1 and 2, we describe four types of
seamless phase 2-3 dose-optimization designs, which are suitable for different
clinical settings. The statistical and design considerations that pertain to
dose optimization are discussed. Simulation shows that dose optimization phase
2-3 designs are able to control the familywise type I error rates and yield
appropriate statistical power with substantially smaller sample size than the
conventional approach. The sample size savings range from 16.6% to 27.3%,
depending on the design and scenario, with a mean savings of 22.1%. Due to the
interim dose selection, the phase 2-3 dose-optimization design is logistically
and operationally more challenging, and should be carefully planned and
implemented to ensure trial integrity.
Authors: Liyun Jiang, Ying Yuan.
We significantly improve performance using properties of the posterior
in our active learning scheme and for the definition of the GP prior. In
particular we account for the expected dynamical range of the posterior in
different dimensionalities. We test our model against a number of synthetic and
cosmological examples.
We present the GPry algorithm for fast Bayesian inference of general
(non-Gaussian) posteriors with a moderate number of parameters. GPry does not
need any pre-training, special hardware such as GPUs, and is intended as a
drop-in replacement for traditional Monte Carlo methods for Bayesian inference.
Our algorithm is based on generating a Gaussian Process surrogate model of the
log-posterior, aided by a Support Vector Machine classifier that excludes
extreme or non-finite values. An active learning scheme allows us to reduce the
number of required posterior evaluations by two orders of magnitude compared to
traditional Monte Carlo inference. Our algorithm allows for parallel
evaluations of the posterior at optimal locations, further reducing wall-clock
times. We significantly improve performance using properties of the posterior
in our active learning scheme and for the definition of the GP prior. In
particular we account for the expected dynamical range of the posterior in
different dimensionalities. We test our model against a number of synthetic and
cosmological examples. GPry outperforms traditional Monte Carlo methods when
the evaluation time of the likelihood (or the calculation of theoretical
observables) is of the order of seconds; for evaluation times of over a minute
it can perform inference in days that would take months using traditional
methods. GPry is distributed as an open source Python package (pip install
gpry) and can also be found at https://github.com/jonaselgammal/GPry.
Authors: Jonas El Gammal, Nils Schöneberg, Jesús Torrado, Christian Fidler.
We consider the fundamental scheduling problem of minimizing the sum of
weighted completion times on a single machine in the non-clairvoyant setting. However, to the best of our knowledge, this concept has never been considered
for the total completion time objective in the non-clairvoyant model. This implies
a performance guarantee of $(1+3\sqrt{3})\approx 6.197$ for the deterministic
algorithm and of $\approx 3.032$ for the randomized version.
We consider the fundamental scheduling problem of minimizing the sum of
weighted completion times on a single machine in the non-clairvoyant setting.
While no non-preemptive algorithm is constant competitive, Motwani, Phillips,
and Torng (SODA '93) proved that the simple preemptive round robin procedure is
$2$-competitive and that no better competitive ratio is possible, initiating a
long line of research focused on preemptive algorithms for generalized variants
of the problem. As an alternative model, Shmoys, Wein, and Williamson (FOCS
'91) introduced kill-and-restart schedules, where running jobs may be killed
and restarted from scratch later, and analyzed then for the makespan objective.
However, to the best of our knowledge, this concept has never been considered
for the total completion time objective in the non-clairvoyant model.
We contribute to both models: First we give for any $b > 1$ a tight analysis
for the natural $b$-scaling kill-and-restart strategy for scheduling jobs
without release dates, as well as for a randomized variant of it. This implies
a performance guarantee of $(1+3\sqrt{3})\approx 6.197$ for the deterministic
algorithm and of $\approx 3.032$ for the randomized version. Second, we show
that the preemptive Weighted Shortest Elapsed Time First (WSETF) rule is
$2$-competitive for jobs released in an online fashion over time, matching the
lower bound by Motwani et al. Using this result as well as the competitiveness
of round robin for multiple machines, we prove performance guarantees of
adaptions of the $b$-scaling algorithm to online release dates and unweighted
jobs on identical parallel machines.
Authors: Sven Jäger, Guillaume Sagnol, Daniel Schmidt genannt Waldschmidt, Philipp Warode.
Frozen pretrained models have become a viable alternative to the
pretraining-then-finetuning paradigm for transfer learning. With this work, we hope to
bring greater attention to this promising path of freezing pretrained image
models.
Frozen pretrained models have become a viable alternative to the
pretraining-then-finetuning paradigm for transfer learning. However, with
frozen models there are relatively few parameters available for adapting to
downstream tasks, which is problematic in computer vision where tasks vary
significantly in input/output format and the type of information that is of
value. In this paper, we present a study of frozen pretrained models when
applied to diverse and representative computer vision tasks, including object
detection, semantic segmentation and video action recognition. From this
empirical analysis, our work answers the questions of what pretraining task
fits best with this frozen setting, how to make the frozen setting more
flexible to various downstream tasks, and the effect of larger model sizes. We
additionally examine the upper bound of performance using a giant frozen
pretrained model with 3 billion parameters (SwinV2-G) and find that it reaches
competitive performance on a varied set of major benchmarks with only one
shared frozen base network: 60.0 box mAP and 52.2 mask mAP on COCO object
detection test-dev, 57.6 val mIoU on ADE20K semantic segmentation, and 81.7
top-1 accuracy on Kinetics-400 action recognition. With this work, we hope to
bring greater attention to this promising path of freezing pretrained image
models.
Authors: Yutong Lin, Ze Liu, Zheng Zhang, Han Hu, Nanning Zheng, Stephen Lin, Yue Cao.
Despite their remarkable performance, deep neural networks remain unadopted
in clinical practice, which is considered to be partially due to their lack in
explainability. We classify data from a public data set and the
attribution methods assign a "relevance score" to each sample of the classified
signals. In summary, our analysis suggests that the DNN learned features similar to
cardiology textbook knowledge.
Despite their remarkable performance, deep neural networks remain unadopted
in clinical practice, which is considered to be partially due to their lack in
explainability. In this work, we apply attribution methods to a pre-trained
deep neural network (DNN) for 12-lead electrocardiography classification to
open this "black box" and understand the relationship between model prediction
and learned features. We classify data from a public data set and the
attribution methods assign a "relevance score" to each sample of the classified
signals. This allows analyzing what the network learned during training, for
which we propose quantitative methods: average relevance scores over a)
classes, b) leads, and c) average beats. The analyses of relevance scores for
atrial fibrillation (AF) and left bundle branch block (LBBB) compared to
healthy controls show that their mean values a) increase with higher
classification probability and correspond to false classifications when around
zero, and b) correspond to clinical recommendations regarding which lead to
consider. Furthermore, c) visible P-waves and concordant T-waves result in
clearly negative relevance scores in AF and LBBB classification, respectively.
In summary, our analysis suggests that the DNN learned features similar to
cardiology textbook knowledge.
Authors: Theresa Bender, Jacqueline Michelle Beinecke, Dagmar Krefting, Carolin Müller, Henning Dathe, Tim Seidler, Nicolai Spicher, Anne-Christin Hauschild.
A high gamma-ray state from the blazar was revealed soon after the event and in a follow-up to about 40 days. A posteriori observations also in the optical and radio bands indicated a rise of the flux from the TXS 0506+056 blazar. This Seyfert II galaxy is at only 14.4~Mpc from the Earth. We discuss these observations.
Neutrino astronomy saw its birth with the discovery by IceCube of a diffuse
flux at energies above 60 TeV with intensity comparable to a predicted upper
limit to the flux from extra-galactic sources of ultra-high energy cosmic rays
(UHECRs). While such an upper limit corresponds to the case of calorimetric
sources, in which cosmic rays lose all their energy into photo-pion production,
the first statistically significant coincident observation between neutrinos
and gamma rays was observed from a blazar of intriguing nature. A
very-high-energy muon event, of most probable neutrino energy of 290 TeV for an
$E^{-2.13}$ spectrum, alerted other observatories triggering a large number of
investigations in many bands of the electromagnetic (EM) spectrum. A high
gamma-ray state from the blazar was revealed soon after the event and in a
follow-up to about 40 days. A posteriori observations also in the optical and
radio bands indicated a rise of the flux from the TXS 0506+056 blazar. A
previous excess of events of the duration of more than 100~d was observed by
IceCube with higher significance than the alert itself. These observations
triggered more complex modeling than simple one-zone proton synchrotron models
for proton acceleration in jets of active galactic nuclei (AGNs) and more
observations across the EM spectrum. A second piece of evidence was a steady
excess of about 50 neutrino events with reconstructed soft spectrum in a sample
of lower energy well-reconstructed muon events than the alert event. A hot spot
was identified in a catalog of 110 gamma-ray intense emitters and starburst
galaxies in a direction compatible with NGC 1068 with a significance of
$2.9\sigma$. NGC 1068 hosts a mildly relativistic jet in a starburst galaxy,
seen not from the jet direction but rather through the torus. This Seyfert II
galaxy is at only 14.4~Mpc from the Earth. We discuss these observations.
Authors: Teresa Montaruli.
Data augmentation leads a remarkable
performance improvement for most of the languages in the inflection task. Our code
https://github.com/emrecanacikgoz/mrl2022 is publicly available.
This paper describes the KUIS-AI NLP team's submission for the 1$^{st}$
Shared Task on Multilingual Clause-level Morphology (MRL2022). We present our
work on all three parts of the shared task: inflection, reinflection, and
analysis. We mainly explore two approaches: Transformer models in combination
with data augmentation, and exploiting the state-of-the-art language modeling
techniques for morphological analysis. Data augmentation leads a remarkable
performance improvement for most of the languages in the inflection task.
Prefix-tuning on pretrained mGPT model helps us to adapt reinflection and
analysis tasks in a low-data setting. Additionally, we used pipeline
architectures using publicly available open source lemmatization tools and
monolingual BERT-based morphological feature classifiers for reinflection and
analysis tasks, respectively. While Transformer architectures with data
augmentation and pipeline architectures achieved the best results for
inflection and reinflection tasks, pipelines and prefix-tuning on mGPT received
the highest results for the analysis task. Our methods achieved first place in
each of the three tasks and outperforms mT5-baseline with ~89\% for inflection,
~80\% for reinflection and ~12\% for analysis. Our code
https://github.com/emrecanacikgoz/mrl2022 is publicly available.
Authors: Emre Can Acikgoz, Tilek Chubakov, Müge Kural, Gözde Gül Şahin, Deniz Yuret.
In this work, we demonstrate coherent electronic control of the photoinduced insulator-to-metal transition in the prototypical Mott insulator V$_2$O$_3$. Comparison between experimental results and numerical solutions of the optical Bloch equations provides an electronic coherence time on the order of 5 fs.
Managing light-matter interaction on timescales faster than the loss of
electronic coherence is key for achieving the full quantum control of final
products in solid-solid transformations. In this work, we demonstrate coherent
electronic control of the photoinduced insulator-to-metal transition in the
prototypical Mott insulator V$_2$O$_3$. Selective excitation of a specific
interband transition with two phase-locked light pulses manipulates the orbital
occupation of the correlated bands in a way that depends on the coherent
evolution of the photoinduced superposition of states. Comparison between
experimental results and numerical solutions of the optical Bloch equations
provides an electronic coherence time on the order of 5 fs. Temperature
dependent experiments suggest that the electronic coherence time is enhanced in
the vicinity of the insulator-to-metal transition critical temperature, thus
highlighting the role of fluctuations in determining the electronic coherence.
These results open new routes to selectively switch functionalities of quantum
materials and coherently control solid-solid electronic transformations.
Authors: Paolo Franceschini, Veronica R. Policht, Alessandra Milloch, Andrea Ronchi, Selene Mor, Simon Mellaerts, Wei-Fan Hsu, Stefania Pagliara, Gabriele Ferrini, Francesco Banfi, Michele Fabrizio, Mariela Menghini, Jean-Pierre Locquet, Stefano Dal Conte, Giulio Cerullo, Claudio Giannetti.
In the single-excitation subspace, this
system not only possesses two energy bands with propagating states, but also
possesses chiral bound states. The
chirality behaviour of the ordinary two bound states outside the energy bands
is quite different from the one of the emerging bound state inside the energy
gap. The almost perfect chiral bound states can be achieved at certain
parameters as a result of completely destructive interference.
We study the chiral feature in a system composed of one two-level emitter
(TLE) and a one dimensional (1D) dimer chain of coupled resonators with the
alternate single-photon energies. In the single-excitation subspace, this
system not only possesses two energy bands with propagating states, but also
possesses chiral bound states. The number of chiral bound states depends on the
coupling forms between the TLE and the dimer chain. It is found that when the
TLE is locally coupled to one resonator of the dimer chain, the bound-state
that has mirror reflection symmetry is not a chiral one. When the TLE is
nonlocally coupled to two adjacent resonators, three chiral bound states arise
due to the mirror symmetry breaking. The chirality of these bound states can be
tuned by changing the energy differences of single photon in the adjacent
resonators, the coupling strengths and the transition energy of the TLE. The
chirality behaviour of the ordinary two bound states outside the energy bands
is quite different from the one of the emerging bound state inside the energy
gap. The almost perfect chiral bound states can be achieved at certain
parameters as a result of completely destructive interference.
Authors: Jing Li, Jing Lu, Z. R. Gong, Lan Zhou.
We find a surprising suppression of the low-energy fluctuations by an external magnetic field at all three dopings. This implies that the response of two-dimensional superconductivity to a magnetic field is similar to that of a bulk superconductor. Our results provide direct evidence of a very gradual onset of superconductivity in cuprates.
We present triple-axis neutron scattering studies of low-energy magnetic
fluctuations in strongly underdoped La$_{2-x}$Sr$_{x}$CuO$_{4}$ with $x=0.05$,
$0.06$ and $0.07$, providing quantitative evidence for a direct competition
between these fluctuations and superconductivity. At dopings $x=0.06$ and
$x=0.07$, three-dimensional superconductivity is found, while only a very weak
signature of two-dimensional superconductivity residing in the CuO$_2$ planes
is detectable for $x=0.05$. We find a surprising suppression of the low-energy
fluctuations by an external magnetic field at all three dopings. This implies
that the response of two-dimensional superconductivity to a magnetic field is
similar to that of a bulk superconductor. Our results provide direct evidence
of a very gradual onset of superconductivity in cuprates.
Authors: Ana-Elena Tutueanu, Machteld E. Kamminga, Tim B. Tejsner, Henrik Jacobsen, Henriette W. Hansen, Monica-Elisabeta Lacatusu, Jacob Baas, Kira L. Eliasen, Jean-Claude Grivel, Yasmine Sassa, Niels Bech Christensen, Paul Steffens, Martin Boehm, Andrea Piovano, Kim Lefmann, Astrid T. Rømer.
Iron ilmenene is a new two-dimensional material that has recently been
exfoliated from the naturally-occurring iron titanate found in ilmenite ore, a
material that is abundant on earth surface. In this work, we theoretically
investigate the structural, electronic and magnetic properties of 2D
transition-metal-based ilmenene-like titanates. Furthermore, the ilmenenes based on late 3d brass metals, such as
CuTiO$_3$ and ZnTiO$_3$, become ferromagnetic and spin compensated,
respectively.
Iron ilmenene is a new two-dimensional material that has recently been
exfoliated from the naturally-occurring iron titanate found in ilmenite ore, a
material that is abundant on earth surface. In this work, we theoretically
investigate the structural, electronic and magnetic properties of 2D
transition-metal-based ilmenene-like titanates. The study of magnetic order
reveals that these ilmenenes usually present intrinsic antiferromagnetic
coupling between the 3d magnetic metals decorating both sides of the Ti-O
layer. Furthermore, the ilmenenes based on late 3d brass metals, such as
CuTiO$_3$ and ZnTiO$_3$, become ferromagnetic and spin compensated,
respectively. Our calculations including spin-orbit coupling reveal that the
magnetic ilmenenes have large magnetocrystalline anisotropy energies when the
3d shell departs from being either filled or half-filled, with their spin
orientation being out-of-plane for elements below half-filling of 3d states and
in-plane above. These interesting magnetic properties of ilmenenes make them
useful for future spintronic applications because they could be synthesized as
already realized in the iron case.
Authors: R. H Aguilera-del-Toro, M. Arruabarrena, A. Leonardo, A. Ayuela.
So data acquisition, processing, communication, and visualization are necessary from a functional standpoint. Sensors capture and covert physical features from their chosen environment into detectable quantities. The received data is interpreted and analyzed with the digital processing circuitry. Ultimately, it is used as information by a network of internet-connected smart devices. Because IoT technologies are adaptable to nearly any technology that may provide its operational activity and environmental conditions. But the challenges occur with power consumption as the complete IoT framework is battery operated and replacing a battery is a daunting task.
Human lives are improving with the widespread use of cutting-edge digital
technology like the Internet of Things (IoT). Recently, the pandemic has shown
the demand for more digitally advanced IoT-based devices. International Data
Corporation (IDC) forecasts that by 2025, there will be approximately 42
billion of these devices in use, capable of producing around 80 ZB (zettabytes)
of data. So data acquisition, processing, communication, and visualization are
necessary from a functional standpoint. Indicating sensors & data converters
are the key components for IoT-based applications. The efficiency of such
applications is truly measured in terms of latency, power, and resolution of
data converters motivating designers to perform efficiently. Sensors capture
and covert physical features from their chosen environment into detectable
quantities. Data converter gives meaningful information and connects the real
analog world to the digital component of the devices. The received data is
interpreted and analyzed with the digital processing circuitry. Ultimately, it
is used as information by a network of internet-connected smart devices.
Because IoT technologies are adaptable to nearly any technology that may
provide its operational activity and environmental conditions. But the
challenges occur with power consumption as the complete IoT framework is
battery operated and replacing a battery is a daunting task. So the goal of
this chapter is to unveil the requirements to design energy-efficient data
converters for IoT applications.
Authors: Buddhi Prakash Sharma, Anu Gupta, Chandra Shekhar.
Recently, DNN-based designs have shown
impressive results in exploiting feedback. However, previous works have focused mainly on passive
feedback, where the transmitter observes a noisy version of the signal at the
receiver. In this work, we show that GBAF codes can also be used for channels
with active feedback.
Deep neural network (DNN)-assisted channel coding designs, such as
low-complexity neural decoders for existing codes, or end-to-end
neural-network-based auto-encoder designs are gaining interest recently due to
their improved performance and flexibility; particularly for communication
scenarios in which high-performing structured code designs do not exist.
Communication in the presence of feedback is one such communication scenario,
and practical code design for feedback channels has remained an open challenge
in coding theory for many decades. Recently, DNN-based designs have shown
impressive results in exploiting feedback. In particular, generalized block
attention feedback (GBAF) codes, which utilizes the popular transformer
architecture, achieved significant improvement in terms of the block error rate
(BLER) performance. However, previous works have focused mainly on passive
feedback, where the transmitter observes a noisy version of the signal at the
receiver. In this work, we show that GBAF codes can also be used for channels
with active feedback. We implement a pair of transformer architectures, at the
transmitter and the receiver, which interact with each other sequentially, and
achieve a new state-of-the-art BLER performance, especially in the low SNR
regime.
Authors: Emre Ozfatura, Yulin Shao, Amin Ghazanfari, Alberto Perotti, Branislav Popovic, Deniz Gunduz.
It is based on a $U(1)$ symmetry in the dark sector, which can be either gauged or global. We discuss the phenomenology of the model and identify the allowed parameter space. We argue that the gauged version of the model is preferred, and in this case the typical energy scale of the model is in the 10 MeV to few GeV range.
Cosmological constraints on the sum of the neutrino masses can be relaxed if
the number density of active neutrinos is reduced compared to the standard
scenario, while at the same time keeping the effective number of neutrino
species $N_{\rm eff}\approx 3$ by introducing a new component of dark
radiation. We discuss a UV complete model to realise this idea, which
simultaneously provides neutrino masses via the seesaw mechanism. It is based
on a $U(1)$ symmetry in the dark sector, which can be either gauged or global.
In addition to heavy seesaw neutrinos, we need to introduce $\mathcal{O}(10)$
generations of massless sterile neutrinos providing the dark radiation. Then we
can accommodate active neutrino masses with $\sum m_\nu \sim 1$ eV, in the
sensitivity range of the KATRIN experiment. We discuss the phenomenology of the
model and identify the allowed parameter space. We argue that the gauged
version of the model is preferred, and in this case the typical energy scale of
the model is in the 10 MeV to few GeV range.
Authors: Miguel Escudero, Thomas Schwetz, Jorge Terol-Calvo.