Search for: All records

Motivated by settings in which predictive models may be required to be non-discriminatory with respect to certain attributes (such as race), but even collecting the sensitive attribute may be forbidden or restricted, we initiate the study of fair learning under the constraint of differential privacy. Our first algorithm is a private implementation of the equalized odds post-processing approach of (Hardt et al., 2016). This algorithm is appealingly simple, but must be able to use protected group membership explicitly at test time, which can be viewed as a form of “disparate treatment”. Our second algorithm is a differentially private version of the oracle-efficient in-processing approach of (Agarwal et al., 2018) which is more complex but need not have access to protected group membership at test time. We identify new tradeoffs between fairness, accuracy, and privacy that emerge only when requiring all three properties, and show that these tradeoffs can be milder if group membership may be used at test time. We conclude with a brief experimental evaluation. 

Asymptotic giant branch stars are responsible for the production of most of the heavy isotopes beyond Sr observed in the solar system. Among them, isotopes shielded from the $r$-process contribution by their stable isobars are defined as $s$-only nuclei. For a long time the abundance of ${}^{204}\mathrm{Pb}$, the heaviest $s$-only isotope, has been a topic of debate because state-of-the-art stellar models appeared to systematically underestimate its solar abundance. Besides the impact of uncertainties from stellar models and galactic chemical evolution simulations, this discrepancy was further obscured by rather divergent theoretical estimates for the neutron capture cross section of its radioactive precursor in the neutron-capture flow, ${}^{204}\mathrm{Tl}$( $t_{1/2}=3.78\mathrm{yr}$), and by the lack of experimental data on this reaction. We present the first ever neutron capture measurement on ${}^{204}\mathrm{Tl}$, conducted at the CERN neutron time-of-flight facility n_TOF, employing a sample of only 9 mg of ${}^{204}\mathrm{Tl}$produced at the Institute Laue Langevin high flux reactor. By complementing our new results with semiempirical calculations we obtained, at the $s$-process temperatures of $kT\approx 8\mathrm{keV}$and $kT\approx 30\mathrm{keV}$, Maxwellian-averaged cross sections (MACS) of 580(168) mb and 260(90) mb, respectively. These figures are about 3% lower and 20% higher than the corresponding values widely used in astrophysical calculations, which were based only on theoretical calculations. By using the new ${}^{204}\mathrm{Tl}$MACS, the uncertainty arising from the ${}^{204}\mathrm{Tl}(\mathrm{n},\gamma )$cross section on the $s$-process abundance of ${}^{204}\mathrm{Pb}$has been reduced from $\sim 30\%$down to $+8\%/-6\%$, and the $s$-process calculations are in agreement with the latest solar system abundance of ${}^{204}\mathrm{Pb}$reported by K. Lodders in 2021. Published by the American Physical Society2024