TY - CONF AB - Data-rich applications in machine-learning and control have motivated an intense research on large-scale optimization. Novel algorithms have been proposed and shown to have optimal convergence rates in terms of iteration counts. However, their practical performance is severely degraded by the cost of exchanging high-dimensional gradient vectors between computing nodes. Several gradient compression heuristics have recently been proposed to reduce communications, but few theoretical results exist that quantify how they impact algorithm convergence. This paper establishes and strengthens the convergence guarantees for gradient descent under a family of gradient compression techniques. For convex optimization problems, we derive admissible step sizes and quantify both the number of iterations and the number of bits that need to be exchanged to reach a target accuracy. Finally, we validate the performance of different gradient compression techniques in simulations. The numerical results highlight the properties of different gradient compression algorithms and confirm that fast convergence with limited information exchange is possible. AU - Khirirat, Sarit AU - Johansson, Mikael AU - Alistarh, Dan-Adrian ID - 7122 SN - 0743-1546 T2 - 2018 IEEE Conference on Decision and Control TI - Gradient compression for communication-limited convex optimization ER - TY - JOUR AB - Prevailing models of sex-chromosome evolution were largely inspired by the stable and highly differentiated XY pairs of model organisms, such as those of mammals and flies. Recent work has uncovered an incredible diversity of sex-determining systems, bringing some of the assumptions of these traditional models into question. One particular question that has arisen is what drives some sex chromosomes to be maintained over millions of years and differentiate fully, while others are replaced by new sex-determining chromosomes before differentiation has occurred. Here, I review recent data on the variability of sex-determining genes and sex chromosomes in different non-model vertebrates and invertebrates, and discuss some theoretical models that have been put forward to account for this diversity. AU - Vicoso, Beatriz ID - 7146 IS - 12 JF - Nature Ecology & Evolution SN - 2397-334X TI - Molecular and evolutionary dynamics of animal sex-chromosome turnover VL - 3 ER - TY - JOUR AB - Roots grow downwards parallel to the gravity vector, to anchor a plant in soil and acquire water and nutrients, using a gravitropic mechanism dependent on the asymmetric distribution of the phytohormone auxin. Recently, Chang et al. demonstrate that asymmetric distribution of another phytohormone, cytokinin, directs root growth towards higher water content. AU - Sinclair, Scott A AU - Friml, Jiří ID - 7143 JF - Cell Research SN - 1001-0602 TI - Defying gravity: a plant's quest for moisture VL - 29 ER - TY - JOUR AB - We propose an efficient microwave-photonic modulator as a resource for stationary entangled microwave-optical fields and develop the theory for deterministic entanglement generation and quantum state transfer in multi-resonant electro-optic systems. The device is based on a single crystal whispering gallery mode resonator integrated into a 3D-microwave cavity. The specific design relies on a new combination of thin-film technology and conventional machining that is optimized for the lowest dissipation rates in the microwave, optical, and mechanical domains. We extract important device properties from finite-element simulations and predict continuous variable entanglement generation rates on the order of a Mebit/s for optical pump powers of only a few tens of microwatts. We compare the quantum state transfer fidelities of coherent, squeezed, and non-Gaussian cat states for both teleportation and direct conversion protocols under realistic conditions. Combining the unique capabilities of circuit quantum electrodynamics with the resilience of fiber optic communication could facilitate long-distance solid-state qubit networks, new methods for quantum signal synthesis, quantum key distribution, and quantum enhanced detection, as well as more power-efficient classical sensing and modulation. AU - Rueda Sanchez, Alfredo R AU - Hease, William J AU - Barzanjeh, Shabir AU - Fink, Johannes M ID - 7156 JF - npj Quantum Information SN - 2056-6387 TI - Electro-optic entanglement source for microwave to telecom quantum state transfer VL - 5 ER - TY - JOUR AB - Cell division, movement and differentiation contribute to pattern formation in developing tissues. This is the case in the vertebrate neural tube, in which neurons differentiate in a characteristic pattern from a highly dynamic proliferating pseudostratified epithelium. To investigate how progenitor proliferation and differentiation affect cell arrangement and growth of the neural tube, we used experimental measurements to develop a mechanical model of the apical surface of the neuroepithelium that incorporates the effect of interkinetic nuclear movement and spatially varying rates of neuronal differentiation. Simulations predict that tissue growth and the shape of lineage-related clones of cells differ with the rate of differentiation. Growth is isotropic in regions of high differentiation, but dorsoventrally biased in regions of low differentiation. This is consistent with experimental observations. The absence of directional signalling in the simulations indicates that global mechanical constraints are sufficient to explain the observed differences in anisotropy. This provides insight into how the tissue growth rate affects cell dynamics and growth anisotropy and opens up possibilities to study the coupling between mechanics, pattern formation and growth in the neural tube. AU - Guerrero, Pilar AU - Perez-Carrasco, Ruben AU - Zagórski, Marcin P AU - Page, David AU - Kicheva, Anna AU - Briscoe, James AU - Page, Karen M. ID - 7165 IS - 23 JF - Development SN - 0950-1991 TI - Neuronal differentiation influences progenitor arrangement in the vertebrate neuroepithelium VL - 146 ER -