Search for: All records

Designing robust loss functions is popular in learning with noisy labels while existing designs did not explicitly consider the overfitting property of deep neural networks (DNNs). As a result, applying these losses may still suffer from overfitting/memorizing noisy labels as training proceeds. In this paper, we first theoretically analyze the memorization effect and show that a lower-capacity model may perform better on noisy datasets. However, it is non-trivial to design a neural network with the best capacity given an arbitrary task. To circumvent this dilemma, instead of changing the model architecture, we decouple DNNs into an encoder followed by a linear classifier and propose to restrict the function space of a DNN by a representation regularizer. Particularly, we require the distance between two self-supervised features to be positively related to the distance between the corresponding two supervised model outputs. Our proposed framework is easily extendable and can incorporate many other robust loss functions to further improve performance. Extensive experiments and theoretical analyses support our claims. Code is available at https://github.com/UCSC-REAL/SelfSup_NoisyLabel. 

Human-annotated labels are often prone to noise, and the presence of such noise will degrade the performance of the resulting deep neural network (DNN) models. Much of the literature (with several recent exceptions) of learning with noisy labels focuses on the case when the label noise is independent of features. Practically, annotations errors tend to be instance-dependent and often depend on the difficulty levels of recognizing a certain task. Applying existing results from instance-independent settings would require a significant amount of estimation of noise rates. Therefore, providing theoretically rigorous solutions for learning with instance-dependent label noise remains a challenge. In this paper, we propose CORES (COnfidence REgularized Sample Sieve), which progressively sieves out corrupted examples. The implementation of CORES does not require specifying noise rates and yet we are able to provide theoretical guarantees of CORES in filtering out the corrupted examples. This high-quality sample sieve allows us to treat clean examples and the corrupted ones separately in training a DNN solution, and such a separation is shown to be advantageous in the instance-dependent noise setting. We demonstrate the performance of CORES^2 on CIFAR10 and CIFAR100 datasets with synthetic instance-dependent label noise and Clothing1M with real-world human noise. As of independent interests, our sample sieve provides a generic machinery for anatomizing noisy datasets and provides a flexible interface for various robust training techniques to further improve the performance. Code is available at https://github.com/UCSC-REAL/cores. 

Self-assembled oxide–metallic alloy nanopillars as hybrid plasmonic metamaterials ( e.g. , ZnO–Ag x Au 1−x ) in a thin film form have been grown using a pulsed laser deposition method. The hybrid films were demonstrated to be highly tunable via systematic tuning of the oxygen background pressure during deposition. The pressure effects on morphology and optical properties have been investigated and found to be critical to the overall properties of the hybrid films. Specifically, low background pressure results in the vertically aligned nanocomposite (VAN) form while the high-pressure results in more lateral growth of the nanoalloys. Strong surface plasmon resonance was observed in the UV-vis region and a hyperbolic dielectric function was achieved due to the anisotropic morphology. The oxide–nanoalloy hybrid material grown in this work presents a highly effective approach for tuning the binary nanoalloy morphology and properties through systematic parametric changes, important for their potential applications in integrated photonics and plasmonics such as sensors, energy harvesting devices, and beyond. 

Sm-doped BiFeO 3 (Bi 0.85 Sm 0.15 FeO 3 , or BSFO) thin films were fabricated on (001) SrTiO 3 (STO) substrates by pulsed laser deposition (PLD) over a range of deposition temperatures (600 °C, 640 °C and 670 °C). Detailed analysis of their microstructure via X-ray diffraction (XRD) and transmission electron microscopy (TEM) shows the deposition temperature dependence of ferroelectric (FE) and antiferroelectric (AFE) phase formation in BSFO. The Sm dopants are clearly detected by high-resolution scanning transmission electron microscopy (HR-STEM) and prove effective in controlling the ferroelectric properties of BSFO. The BSFO ( T dep = 670 °C) presents larger remnant polarization (Pr) than the other two BSFO ( T dep = 600 °C, 640 °C) and pure BiFeO 3 (BFO) thin films. This study paves a simple way for enhancing the ferroelectric properties of BSFO via deposition temperature and further promoting BFO practical applications. 

The emerging field of self-assembled vertically aligned nanocomposite (VAN) thin films effectively enables strain, interface, and microstructure engineering as well as (multi)functional improvements in electric, magnetic, optical, and energy-related properties. Well-ordered or patterned microstructures not only empower VAN thin films with many new functionalities but also enable VAN thin films to be used in nanoscale devices. Comparative ordered devices formed via templating methods suffer from critical drawbacks of processing complexity and potential contamination. Therefore, VAN thin films with spontaneous ordering stand out and display many appealing features for next-generation technological devices, such as electronics, optoelectronics, ultrahigh-density memory systems, photonics, and 3D microbatteries. The spontaneous ordering described in this review contains ordered/patterned structures in both in-plane and out-of-plane directions. In particular, approaches to obtaining spontaneously ordered/patterned structures in-plane are systematically reviewedfrom both thermodynamic and kinetic perspectives. Out-of-plane ordering is also discussed in detail. In addition to reviewing the progress of VAN films with spontaneous ordering, this article also highlights some recent developments in spontaneous ordering approaches and proposes future directions.