arXiv subjects

Bingjing Zhang

Publications and source records attributed to Bingjing Zhang.

3 recordsLinked to original sources

arXiv2018

Fault Tolerance in Iterative-Convergent Machine Learning

Machine learning (ML) training algorithms often possess an inherent self-correcting behavior due to their iterative-convergent nature. Recent systems exploit this property to achieve adaptability and efficiency in unreliable computing environments by relaxing the consistency of execution and allowing calculation errors to be self-corrected during training. However, the behavior of such systems are only well understood for specific types of calculation errors, such as those caused by staleness, reduced precision, or asynchronicity, and for specific types of training algorithms, such as stochastic gradient descent. In this paper, we develop a general framework to quantify the effects of calculation errors on iterative-convergent algorithms and use this framework to design new strategies for checkpoint-based fault tolerance. Our framework yields a worst-case upper bound on the iteration cost of arbitrary perturbations to model parameters during training. Our system, SCAR, employs strategies which reduce the iteration cost upper bound due to perturbations incurred when recovering from checkpoints. We show that SCAR can reduce the iteration cost of partial failures by 78% - 95% when compared with traditional checkpoint-based fault tolerance across a variety of ML models and training algorithms.

cs.LG↗

arXiv2018

Giant and anisotropic many-body spin-orbit tunability in a strongly correlated kagome magnet

Owing to the unusual geometry of kagome lattices-lattices made of corner-sharing triangles-their electrons are useful for studying the physics of frustrated, correlated and topological quantum electronic states. In the presence of strong spin-orbit coupling, the magnetic and electronic structures of kagome lattices are further entangled, which can lead to hitherto unknown spin-orbit phenomena. Here we use a combination of vector-magnetic-field capability and scanning tunnelling microscopy to elucidate the spin-orbit nature of the kagome ferromagnet Fe3Sn2 and explore the associated exotic correlated phenomena. We discover that a many-body electronic state from the kagome lattice couples strongly to the vector field with three-dimensional anisotropy, exhibiting a magnetization-driven giant nematic (two-fold-symmetric) energy shift. Probing the fermionic quasi-particle interference reveals consistent spontaneous nematicity-a clear indication of electron correlation-and vector magnetization is capable of altering this state, thus controlling the many-body electronic symmetry. These spin-driven giant electronic responses go well beyond Zeeman physics and point to the realization of an underlying correlated magnetic topological phase. The tunability of this kagome magnet reveals a strong interplay between an externally applied field, electronic excitations and nematicity, providing new ways of controlling spin-orbit properties and exploring emergent phenomena in topological or quantum materials.

cond-mat.str-el↗

arXiv2017

A Hybrid Supervised-unsupervised Method on Image Topic Visualization with Convolutional Neural Network and LDA

Given the progress in image recognition with recent data driven paradigms, it's still expensive to manually label a large training data to fit a convolutional neural network (CNN) model. This paper proposes a hybrid supervised-unsupervised method combining a pre-trained AlexNet with Latent Dirichlet Allocation (LDA) to extract image topics from both an unlabeled life-logging dataset and the COCO dataset. We generate the bag-of-words representations of an egocentric dataset from the softmax layer of AlexNet and use LDA to visualize the subject's living genre with duplicated images. We use a subset of COCO on 4 categories as ground truth, and define consistent rate to quantitatively analyze the performance of the method, it achieves 84% for consistent rate on average comparing to 18.75% from a raw CNN model. The method is capable of detecting false labels and multi-labels from COCO dataset. For scalability test, parallelization experiments are conducted with Harp-LDA on a Intel Knights Landing cluster: to extract 1,000 topic assignments for 241,035 COCO images, it takes 10 minutes with 60 threads.

cs.CV↗