Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userWhich Model to Transfer? Finding the Needle in the Growing Haystack
123
0
0.0
(
0
)
Added by
Cedric Renggli
Publication date
2020
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Transfer learning has been recently popularized as a data-efficient alternative to training models from scratch, in particular in vision and NLP where it provides a remarkably solid baseline. The emergence of rich model repositories, such as TensorFlow Hub, enables the practitioners and researchers to unleash the potential of these models across a wide range of downstream tasks. As these repositories keep growing exponentially, efficiently selecting a good model for the task at hand becomes paramount. We provide a formalization of this problem through a familiar notion of regret and introduce the predominant strategies, namely task-agnostic (e.g. picking the highest scoring ImageNet model) and task-aware search strategies (such as linear or kNN evaluation). We conduct a large-scale empirical study and show that both task-agnostic and task-aware methods can yield high regret. We then propose a simple and computationally efficient hybrid search strategy which outperforms the existing approaches. We highlight the practical benefits of the proposed solution on a set of 19 diverse vision tasks.
rate research
Read More
Important tasks like record linkage and extreme classification demonstrate extreme class imbalance, with 1 minority instance to every 1 million or more majority instances. Obtaining a sufficient sample of all classes, even just to achieve statistically-significant evaluation, is so challenging that most current approaches yield poor estimates or incur impractical cost. Where importance sampling has been levied against this challenge, restrictive constraints are placed on performance metrics, estimates do not come with appropriate guarantees, or evaluations cannot adapt to incoming labels. This paper develops a framework for online evaluation based on adaptive importance sampling. Given a target performance metric and model for $p(y|x)$, the framework adapts a distribution over items to label in order to maximize statistical precision. We establish strong consistency and a central limit theorem for the resulting performance estimates, and instantiate our framework with worked examples that leverage Dirichlet-tree models. Experiments demonstrate an average MSE superior to state-of-the-art on fixed label budgets.
Context: Isolated cooling neutron stars with thermal X-ray emission remain rarely detected objects despite many searches investigating the ROSAT data. Aims: We simulate the population of close-by young cooling neutron stars to explain the current observational results. Given the inhomogeneity of the neutron star distribution on the sky it is particularly interesting to identify promising sky regions with respect to on-going and future searches. Methods: Applying a population synthesis model the inhomogeneity of the progenitor distribution and the inhomogeneity of the X-ray absorbing interstellar medium are considered for the first time. The total number of observable neutron stars is derived with respect to ROSAT count rates. In addition, we present sky maps of neutron star locations and discuss age and distance distributions of the simulated neutron stars. Implications for future searches are discussed. Results: With our advanced model we can successfully explain the observed logN - logS distribution of close-by neutron stars. Cooling neutron stars will be most abundant in the directions of rich OB associations. New candidates are expected to be identified behind the Gould Belt, in particular in the Cygnus-Cepheus region. They are expected to be on average younger and then hotter than the known population of isolated cooling neutron stars. In addition, we propose to use data on runaway stars to search for more radio-quiet cooling neutron stars.
The Fermi Gamma-ray Space Telescope has greatly expanded the number and energy window of observations of gamma-ray bursts (GRBs). However, the coarse localizations of tens to a hundred square degrees provided by the Fermi GRB Monitor instrument have posed a formidable obstacle to locating the bursts host galaxies, measuring their redshifts, and tracking their panchromatic afterglows. We have built a target-of-opportunity mode for the intermediate Palomar Transient Factory in order to perform targeted searches for Fermi afterglows. Here, we present the results of one year of this program: 8 afterglow discoveries out of 35 searches. Two of the bursts with detected afterglows (GRBs 130702A and 140606B) were at low redshift (z=0.145 and 0.384 respectively) and had spectroscopically confirmed broad-line Type Ic supernovae. We present our broadband follow-up including spectroscopy as well as X-ray, UV, optical, millimeter, and radio observations. We study possible selection effects in the context of the total Fermi and Swift GRB samples. We identify one new outlier on the Amati relation. We find that two bursts are consistent with a mildly relativistic shock breaking out from the progenitor star, rather than the ultra-relativistic internal shock mechanism that powers standard cosmological bursts. Finally, in the context of the Zwicky Transient Facility, we discuss how we will continue to expand this effort to find optical counterparts of binary neutron star mergers that may soon be detected by Advanced LIGO and Virgo.
Network device syslogs are ubiquitous and abundant in modern data centers with most large data centers producing millions of messages per day. Yet, the operational information reflected in syslogs and their implications on diagnosis or management tasks are poorly understood. Prevalent approaches to understanding syslogs focus on simple correlation and abnormality detection and are often limited to detection providing little insight towards diagnosis and resolution. Towards improving data center operations, we propose and implement Log-Prophet, a system that applies a toolbox of statistical techniques and domain-specific models to mine detailed diagnoses. Log-Prophet infers causal relationships between syslog lines and constructs succinct but valuable problem graphs, summarizing root causes and their locality, including cascading problems. We validate Log-Prophet using problem tickets and through operator interviews. To demonstrate the strength of Log-Prophet, we perform an initial longitudinal study of a large online service providers data center. Our study demonstrates that Log-Prophet significantly reduces the number of alerts while highlighting interesting operational issues.
We explore large-$N$ symmetric orbifolds of the $mathcal N=2$ minimal models, and find evidence that their moduli spaces each contain a supergravity point. We identify single-trace exactly marginal operators that deform them away from the symmetric orbifold locus. We also show that their elliptic genera exhibit slow growth consistent with supergravity spectra in AdS$_3$. We thus propose an infinite family of new holographic CFTs.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
