Search for: All records

We give the first provably efficient algorithms for learning neural networks with distribution shift. We work in the Testable Learning with Distribution Shift framework (TDS learning) of Klivans et al. (2024), where the learner receives labeled examples from a training distribution and unlabeled examples from a test distribution and must either output a hypothesis with low test error or reject if distribution shift is detected. No assumptions are made on the test distribution. All prior work in TDS learning focuses on classification, while here we must handle the setting of nonconvex regression. Our results apply to real-valued networks with arbitrary Lipschitz activations and work whenever the training distribution has strictly sub-exponential tails. For training distributions that are bounded and hypercontractive, we give a fully polynomial-time algorithm for TDS learning one hidden-layer networks with sigmoid activations. We achieve this by importing classical kernel methods into the TDS framework using data-dependent feature maps and a type of kernel matrix that couples samples from both train and test distributions. 

Abstract Human induced pluripotent stem cell-derived (iPSC) neural cultures offer clinically relevant models of human diseases, including Amyotrophic Lateral Sclerosis, Alzheimer’s, and Autism Spectrum Disorder. In situ characterization of the spatial-temporal evolution of cell state in 3D culture and subsequent 2D dissociated culture models based on protein expression levels and localizations is essential to understanding neural cell differentiation, disease state phenotypes, and sample-to-sample variability. Here, we apply PR obe-based I maging for S equential M ultiplexing (PRISM) to facilitate multiplexed imaging with facile, rapid exchange of imaging probes to analyze iPSC-derived cortical and motor neuron cultures that are relevant to psychiatric and neurodegenerative disease models, using over ten protein targets. Our approach permits analysis of cell differentiation, cell composition, and functional marker expression in complex stem-cell derived neural cultures. Furthermore, our approach is amenable to automation, offering in principle the ability to scale-up to dozens of protein targets and samples. 

A new technique is developed to identify dielectrons ( $e^{+}{e}^{-}$ ) with Lorentz boost ${\gamma }_{\mathrm{L}}>20$ that produce one single merged cluster in the electromagnetic calorimeter of the CMS detector. The identification uses two multivariate models: one for the case where both electron tracks are reconstructed, and another where only one of the tracks is reconstructed. The efficiency is determined using proton-proton collision data collected at a center-of-mass energy of 13 TeV. Boosted $J/\psi$ mesons decaying into $e^{+}{e}^{-}$ pairs are used to estimate the efficiency of the model with two tracks, yielding an overall efficiency of 80%. The $Z\to {\mu }^{+}{\mu }^{-}\gamma$ events, where the photon converts into a collimated dielectron, are used for the model with a single track, yielding an efficiency of about 60%. A dedicated energy correction for dielectron candidates is also developed using $B^{\pm }\to J/\psi K^{\pm }\to e^{+}{e}^{-}{K}^{\pm }$ data. 

A search for Higgs boson ( $H$ ) production in association with two vector bosons ( $V=W$ , $Z$ ) via vector-boson scattering (VBS) is presented using proton-proton collision data collected at $\sqrt{s}=13\mathrm{TeV}$ by the CMS experiment, corresponding to an integrated luminosity of $138{\mathrm{fb}}^{-1}$ . Events containing two forward jets consistent with VBS, a large-radius jet from the decay of a boosted $H$ to a pair of $b$ quarks, and zero, one, or two charged leptons coming from $V$ decays are selected. The process is excluded at 95% confidence level for observed (expected) values of the $VVHH$ coupling modifier ${\kappa }_{2V}$ outside the interval $0.40<{\kappa }_{2V}<1.60$ ( $0.34<{\kappa }_{2V}<1.66$ ), assuming standard model values for all other couplings, thus establishing a novel probe of the $VVHH$ interaction. Constraints are also set on the individual ${\kappa }_{2W}$ and ${\kappa }_{2Z}$ coupling modifiers and on the allowed region in the ${\kappa }_{2W}\text{−}{\kappa }_{2Z}$ plane. 

Background Hypertension drives myocardial remodeling, leading to changes in structure, composition and mechanical behavior, including residual stress, which are linked to heart disease progression in a gender-specific manner. Emerging therapies are also targeting constituent-specific pathological features. All previous studies, however, have characterized remodeling in the intact tissue, rather than isolated tissue constituents, and did not include sex as a biological variable. Objective In this study we first identified the contribution of collagen fiber network and myocytes to the myocardial residual stress/strain in Dahl-Salt sensitive rats fed with high fat diet. Then, we quantified the effect of hypertension on the remodeling of the left ventricle (LV), as well as the existence of sex-specific remodeling features. Methods We performed mechanical tests (opening angle, ring-test) and histological analysis on isolated constituents and intact tissue of the LV. Based on the measurements from the tests, we performed a stress analysis to evaluate the residual stress distribution. Statistical analysis was performed to identify the effects of constituent isolation, elevated blood pressure, and sex of the animal on the experimental measurements and modeling results. Results Hypertension leads to reduced residual stress/strain in the intact tissue, isolated collagen fibers, and isolated myocytes in male and female rats. Collagen remains the largest contributor to myocardial residual stress in both normotensive and hypertensive animals. We identified sex-differences in both hypertensive and normotensive animals. Conclusions We observed both constituent- and sex-specific remodeling features in the LV of an animal model of hypertension. 

Evidence of the production of two $Z$ bosons and a photon in proton-proton collisions is reported for the first time in CMS. The analysis uses data collected by the CMS experiment between 2016 and 2018 at a center-of-mass energy of 13 TeV, corresponding to an integrated luminosity of $138{\mathrm{fb}}^{-1}$ . Evidence for the process $pp\to ZZ\gamma \to 4\ell \gamma$ ( $\ell =e$ , $\mu$ ), with an observed (expected) significance of 3.7 (3.1) standard deviations in a fiducial region defined by $p_{\mathrm{T}}^{\gamma }>20\mathrm{GeV}$ , $\left|{\eta }^{\gamma }\right|<2.4$ , $\mathrm{\Delta }R(\ell ,\gamma )>0.5$ , $m_Z$ between 60 and 120 GeV, and the invariant mass of either of the two $Z$ bosons combined with the photon ( $m_{Z_{\gamma }}$ ) larger than 100 GeV, is reported. The measured (predicted) fiducial cross section is ${60}_{-22}^{+27}\mathrm{ab}$ ( $47.56\pm 0.04\mathrm{ab}$ ). Additionally, the inclusive production of $pp\to 4\ell \gamma$ is studied by removing the $m_{Z_{\gamma }}$ requirement to include final state radiation where one $Z$ boson decays to $2\ell \gamma$ , yielding an observed (expected) significance of 5.0 (4.2) standard deviations and a measured (predicted) fiducial cross section of ${156}_{-35}^{+39}\mathrm{ab}$ ( $99.97\pm 0.09\mathrm{ab}$ ).