Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (6 results)

  • Thesis
  • Peer Reviewed
UC San Diego Electronic Theses and Dissertations (2021)

Modern data science calls for statistical inference algorithms that are both data-efficient and computation-efficient. We design and analyze methods that 1) outperform existing algorithms in the high-dimensional setting; 2) are as simple as possible but efficiently computable.

One line of our work aims to bridge different fields and offer unified solutions to multiple questions. Consider the following canonical statistical inference tasks: distribution estimation, functional estimation, and property testing, sharing the model that provides sample access to an unknown discrete distribution. In a recent paper in NeurIPS '19, we showed that a single, simple, and unified estimator – profile maximum likelihood (PML), and its near-linear time computable variant are sample-optimal for estimating multiple distribution attributes. The result covers 1) any appropriately Lipschitz and additively separable functionals; 2) sorted distribution; 3) R enyi entropy; 4) $\ell_2$ distance to the uniform distribution, yielding an optimal tester for distributions' closeness. This work makes PML the first unified sample- and time-optimal method for the learning tasks mentioned above. A single algorithm with such broad applicability is \emph{universal}.

Another line of our work focuses on instance-optimal learning that designs algorithms with near-optimal guarantees for every possible data input. A flagship problem is distribution estimation over discrete or continuous domains, where ordering and geometry play an essential role. Going beyond worst-case guarantees, researchers designed algorithms that compete with a genie estimator that knows the actual distribution but is reasonably restricted. To obtain state-of-the-art algorithms for both tasks, we leveraged the simple but nontrivial idea of ``data binning''. For discrete settings, we group symbols that appear the same number of times. And for continuous settings, we partition the real domain and separate symbols according to pre-designed local quantiles. The respective algorithms run in near-linear-time, achieve the best-known estimation guarantees regarding the genie estimators, and appear in ICML'19 and NeurIPS '20. A genie-like algorithm adaptive to almost every data sample is \emph{competitive}.

We present a comprehensive understanding of universal and competitive algorithms for multiple fundamental learning problems. Our ideas and techniques may shed light on key challenges in modern data science and numerous applications beyond the scope of this dissertation.

    Cover page: Competitive and Universal Learning
    • Article
    • Peer Reviewed
    Chemical Sciences (2015)
    Zinc sulfide nanosheet-based hybrid superlattices with tunable periodicities and rod-like or tubular morphologies are constructed by the spontaneous assembly of nanosheets as they grow in amine solution. The as-prepared architectures can be used as an enhanced electrode for photocurrent response and converted to other functional materials.
      Cover page: Zinc Sulfide Nanosheet‐Based Hybrid Superlattices with Tunable Architectures Showing Enhanced Photoelectrochemical Properties
      • Article
      • Peer Reviewed
      LBL Publications (2016)
      Biogenetic nanopalladium (bio-Pd) has attracted increasing attention recently due to its economical and environmental friendly synthesis route. However, traditional bacteria suspensions formed palladium (Sus-Pd) is limited to be the electrochemical catalyst owing to the poor conductivity of bacterial cells. Herein, we demonstrated Pd nanoparticles, synthesized by electroactive Geobacter biofilm, can form a three-dimensional conductive network (EAB-Pd) that is beneficial to the electrons transfer. As a result, the EAB-Pd delivered an over 5-fold increase of current compared to the Sus-Pd in hydrogen evolution and the reductive degradation of nitro-, azo- and chloroaromatics. Superior performance of EAB-Pd was also observed in comparison with the commercial Pd catalyst. A good stability of EAB-Pd has been further confirmed under electrochemical and mechanical stresses as well as through the reuse after over 3 months of storage. This novel proposed method enables the direct electrochemical application of bio-Pd without the previous required cell carbonization and chemical binders.
        Cover page: Electroactive Biofilm Serving as the Green Synthesizer and Stabilizer for in Situ Fabricating 3D Nanopalladium Network: An Efficient Electrocatalyst
        • Article
        • Peer Reviewed
        LBL Publications (2017)
        The palladized cell (Pd-cell) could be used as an efficient catalyst in catalyzing the degradations of a wide variety of environmental contaminants. Nevertheless, when the Pd NPs associate with the bacteria, the catalytic activity likely significantly affected by the biomass. Quantitative indicators that characterize of Pd-cell are necessary and little attention has been paid to investigate how the catalytic efficiency of Pd-cell is affected by the size and distribution of Pd NPs. To fill this gap, we explored the roles of the above-mentioned key factors on the performance of Pd-cell in catalyzing the degradations of two aromatic contaminants (nitrobenzene and p-chlorophenol) in two commonly used scenarios: (1) using Pd-cell as suspended catalyst in solution and (2) using Pd-cell as electrocatalyst directly coated on electrode. In scenario (1), the relationship of exposing area to Pd particle size and distribution factors was established. Based on theoretical estimation and catalytic performance analysis, the results indicated that adjusting the exposing area to a large value (9.3 ± 0.1 × 105 nm2 mg-1 Pd) was extremely effective for improving the catalytic activity of Pd-cell used as a suspension catalyst. In scenario (2), our results showed that the best electrocatalytic performances were achieved on the electrode decorated with Pd-cells with the largest NP size (54.3 ± 16.4 nm), which exerted maximum electrochemical active surface area (10.6 m2 g-1) as well as favorable conductivity. The coverage of deposited Pd NPs (>95%) on the cell surface played a crucial role in boosting the conductivity of biocatalyst, thus determining the possibility of Pd-cell as an efficient electrocatalyst. The findings of this study provide a guidance for the synthesis and application of Pd-cell, which enables the design of Pd-cell to be suitable for different catalysis systems with high catalytic performance.
          Cover page: Palladized cells as suspension catalyst and electrochemical catalyst for reductively degrading aromatics contaminants: Roles of Pd size and distribution
          • Article
          • Peer Reviewed
          UCLA Previously Published Works (2020)
          Energy circulation in geospace lies at the heart of space weather research. In the inner magnetosphere, the steep plasmapause boundary separates the cold dense plasmasphere, which corotates with the planet, from the hot ring current/plasma sheet outside. Theoretical studies suggested that plasmapause surface waves related to the sharp inhomogeneity exist and act as a source of geomagnetic pulsations, but direct evidence of the waves and their role in magnetospheric dynamics have not yet been detected. Here, we show direct observations of a plasmapause surface wave and its impacts during a geomagnetic storm using multi-satellite and ground-based measurements. The wave oscillates the plasmapause in the afternoon-dusk sector, triggers sawtooth auroral displays, and drives outward-propagating ultra-low frequency waves. We also show that the surface-wave-driven sawtooth auroras occurred in more than 90% of geomagnetic storms during 2014-2018, indicating that they are a systematic and crucial process in driving space energy dissipation.
            Cover page: Plasmapause surface wave oscillates the magnetosphere and diffuse aurora
            • Article
            • Peer Reviewed
            LBL Publications (2020)
            We measure the projected number density profiles of galaxies and the splashback feature in clusters selected by the Sunyaev--Zeldovich (SZ) effect from the Advanced Atacama Cosmology Telescope (AdvACT) survey using galaxies observed by the Dark Energy Survey (DES). The splashback radius for the complete galaxy sample is consistent with theoretical measurements from CDM-only simulations, and is located at $2.4^{+0.3}_{-0.4}$ Mpc $h^{-1}$. We split the sample based on galaxy color and find significant differences in the profile shapes. Red galaxies and those in the green valley show a splashback-like minimum in their slope profile consistent with theoretical predictions, while the bluest galaxies show a weak feature that appears at a smaller radius. We develop a mapping of galaxies to subhalos in $N$-body simulations by splitting subhalos based on infall time onto the cluster halos. We find that the location of the steepest slope and differences in the shapes of the profiles can be mapped to differences in the average time of infall of galaxies of different colors. The minima of the slope in the galaxy profiles trace a discontinuity in the phase space of dark matter halos. By relating spatial profiles to infall time for galaxies of different colours, we can use splashback as a clock to understand galaxy quenching. We find that red galaxies have on average been in their clusters for over $3.2 ~\rm Gyrs$, green galaxies about $2.2 ~\rm Gyrs$, while blue galaxies have been accreted most recently and have not reached apocenter. Using the information from the complete radial profiles, we fit a simple quenching model and find that the onset of galaxy quenching in clusters occurs after a delay of about a gigayear, and that galaxies quench rapidly thereafter with an exponential timescale of $0.6$ Gyr.
              Top