Search

Arbitrary, inconsistent, or faulty decision-making raises serious concerns, and preventing unfair models is an increasingly important challenge in Machine Learning. Data often reflect past discriminatory behavior, and models trained on such data may reflect bias on sensitive attributes, such as gender, race, or age. One approach to developing fair models is to preprocess the training data to remove the underlying biases while preserving the relevant information, for example, by correcting biased labels. While multiple label noise correction methods are available, the information about their behavior in identifying discrimination is very limited. In this work, we develop an empirical methodology to systematically evaluate the effectiveness of label noise correction techniques in ensuring the fairness of models trained on biased datasets. Our methodology involves manipulating the amount of label noise and can be used with fairness benchmarks but also with standard ML datasets. We apply the methodology to analyze six label noise correction methods according to several fairness metrics on standard OpenML datasets. Our results suggest that the Hybrid Label Noise Correction method achieves the best trade-off between predictive performance and fairness. Clustering-Based Correction can reduce discrimination the most, however, at the cost of lower predictive performance.

We experimentally demonstrate stable trapping and controlled manipulation of silica microspheres in a structured optical beam consisting of a dark focus surrounded by light in all directions - the so-called Dark Focus Tweezer. Results from power spectrum and potential analysis demonstrate the non-harmonicity of the trapping potential landspace, which is reconstructed from experimental data in agreement to Lorentz-Mie numerical simulations. Applications of the dark tweezer in levitated optomechanics and biophysics are discussed., Comment: Final version

The goal of this paper is to increase the membership list of the Chamaeleon star forming region and the $\epsilon$ Cha moving group, in particular for low-mass stars and substellar objects. We extended the search region significantly beyond the dark clouds. Our sample has been selected based on proper motions and colours obtained from Gaia and 2MASS. We present and discuss the optical spectroscopic follow-up of 18 low-mass stellar objects in Cha I and $\epsilon$ Cha. We characterize the properties of objects by deriving their physical parameters, both from spectroscopy and photometry. We add three more low-mass members to the list of Cha I, and increase the census of known $\epsilon$ Cha members by more than 40%, confirming spectroscopically 13 new members and relying on X-ray emission as youth indicator for 2 more. In most cases the best-fitting spectral template is from objects in the TW Hya association, indicating that $\epsilon$ Cha has a similar age. The first estimate of the slope of the initial mass function in $\epsilon$ Cha down to the sub-stellar regime is consistent with that of other young clusters. We estimate our IMF to be complete down to $\approx 0.03$M$_{\odot}$. The IMF can be represented by two power laws: for M $<$ 0.5 M$_{\odot}$ $\alpha = 0.42 \pm 0.11$ and for M $>$ 0.5 M$_{\odot}$ $\alpha = 1.44 \pm 0.12$. We find similarities between $\epsilon$ Cha and the southernmost part of Lower Centaurus Crux (LCC A0), both lying at similar distances and sharing the same proper motions. This suggests that $\epsilon$ Cha and LCC A0 may have been born during the same star formation event, Comment: 18 pages, 15 figure, accepted by A&A

We develop a microscopic toy model for Cavity dynamical Casimir effect (DCE), namely, the photon generation from vacuum due to a nonstationary dielectric slab in a fixed single mode cavity. We represent the slab by $N\gg 1$ noninteracting two-level atoms coupled to the field via the standard dipole interaction. We show that the DCE is contained implicitly in the light-matter interaction Hamiltonian when its parameters are externally prescribed functions of time. We also predict several new phenomena, such as saturation of the photon growth due to effective Kerr nonlinearity, generation of pairs of atomic excitations instead of photons ("Inverse DCE") and coherent annihilation of pair of system excitations due to the atomic modulation ("Anti-DCE"). These results are extended to the circuit QED architecture, where similar effects can be implemented with a single qubit providing an alternative way to generate cavity and atom-field entangled states., Comment: 27 pages