skip to main content

Search for: All records

Creators/Authors contains: "Zhang, Fan"

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Abstract

    Numerical solutions to the Einstein constraint equations are constructed on a selection of compact orientable three-dimensional manifolds with non-trivial topologies. A simple constant mean curvature solution and a somewhat more complicated non-constant mean curvature solution are computed on example manifolds from three of the eight Thursten geometrization classes. The constant mean curvature solutions found here are also solutions to the Yamabe problem that transforms a geometry into one with constant scalar curvature.

  2. Abstract

    Understanding human mobility is of great significance for sustainable transportation planning. Long-term travel delay change is a key metric to measure human mobility evolution in cities. However, it is challenging to quantify the long-term travel delay because it happens in different modalities, e.g., subway, taxi, bus, and personal cars, with implicated coupling. More importantly, the data for long-term multi-modal delay modeling is challenging to obtain in practice. As a result, the existing travel delay measurements mainly focus on either single-modal system or short-term mobility patterns, which cannot reveal the long-term travel dynamics and the impact among multi-modal systems. In this paper, we perform a travel delay measurement study to quantify and understand long-term multi-modal travel delay. Our measurement study utilizes a 5-year dataset of 8 million residents from 2013 to 2017 including a subway system with 3 million daily passengers, a 15 thousand taxi system, a 10 thousand personal car system, and a 13 thousand bus system in the Chinese city Shenzhen. We share new observations as follows: (1) the aboveground system has a higher delay increase overall than that of the underground system but the increase of it is slow down; (2) the underground system infrastructure upgrades decreasesmore »the aboveground system travel delay increase in contrast to the increase the underground system travel delay caused by the aboveground system infrastructure upgrades; (3) the travel delays of the underground system decreases in the higher population region and during the peak hours.

    « less
  3. In-Memory Computing (IMC) technology has been considered to be a promising approach to solve well-known memory-wall challenge for data intensive applications. In this paper, we are the first to propose MnM, a novel IMC system with innovative architecture/circuit designs for fast and efficient Min/Max searching computation in emerging Spin-Orbit Torque Magnetic Random Access Memory (SOT-MRAM). Our proposed SOT-MRAM based in-memory logic circuits are specially optimized to perform parallel, one-cycle XNOR logic that are heavily used in the Min/Max searching-in-memory algorithm. Our novel in-memory XNOR circuit also has an overhead of just two transistors per row when compared to most prior methodologies which typically use multiple sense amplifiers or complex CMOS logic gates. We also design all other required peripheral circuits for implementing complete Min/Max searching-in-MRAM computation. Our cross-layer comprehensive experiments on Dijkstra's algorithm and other sorting algorithms in real word datasets show that our MnM could achieve significant performance improvement over CPUs, GPUs, and other competing IMC platforms based on RRAM/MRAM/DRAM.
    Free, publicly-accessible full text available June 6, 2023
  4. Free, publicly-accessible full text available August 11, 2023
  5. ReRAM crossbar array as a high-parallel fast and energy-efficient structure attracts much attention, especially on the acceleration of Deep Neural Network (DNN) inference on one specific task. However, due to the high energy consumption of weight re-programming and the ReRAM cells’ low endurance problem, adapting the crossbar array for multiple tasks has not been well explored. In this paper, we propose XMA, a novel crossbar-aware shift-based mask learning method for multiple task adaption in the ReRAM crossbar DNN accelerator for the first time. XMA leverages the popular mask-based learning algorithm’s benefit to mitigate catastrophic forgetting and learn a task-specific, crossbar column-wise, and shift-based multi-level mask, rather than the most commonly used elementwise binary mask, for each new task based on a frozen backbone model. With our crossbar-aware design innovation, the required masking operation to adapt for a new task could be implemented in an existing crossbar-based convolution engine with minimal hardware/memory overhead and, more importantly, no need for power-hungry cell re-programming, unlike prior works. The extensive experimental results show that, compared with state-of-the art multiple task adaption Piggyback method [1], XMA achieves 3.19% higher accuracy on average, while saving 96.6% memory overhead. Moreover, by eliminating cell re-programming, XMA achieves ∼4.3×more »higher energy efficiency than Piggyback.« less
    Free, publicly-accessible full text available July 10, 2023
  6. Free, publicly-accessible full text available August 10, 2023
  7. Free, publicly-accessible full text available April 14, 2023
  8. Free, publicly-accessible full text available July 1, 2023
  9. Leveraging the ReRAM crossbar-based In-Memory-Computing (IMC) to accelerate single task DNN inference has been widely studied. However, using the ReRAM crossbar for continual learning has not been explored yet. In this work, we propose XST, a novel crossbar column-wise sparse training framework for continual learning. XST significantly reduces the training cost and saves inference energy. More importantly, it is friendly to existing crossbar-based convolution engine with almost no hardware overhead. Compared with the state-of-the-art CPG method, the experiments show that XST's accuracy achieves 4.95 % higher accuracy. Furthermore, XST demonstrates ~5.59 × training speedup and 1.5 × inference energy-saving.
    Free, publicly-accessible full text available March 14, 2023
  10. This paper studies MIMO relays with non-identical link coherence times, a frequently occurring condition when, e.g., the nodes in the relay channel do not all have the same mobility, or the scatterers around some nodes have different mobility compared with those around other nodes. Despite its practical relevance, this condition, known as coherence diversity, has not been studied in the relay channel. This paper studies the performance of MIMO relays and proposes efficient transmission strategies under coherence diversity. Since coherence times have a prominent impact on channel training, we do not assume channel state is available to the decoder for free; all channel training resources are accounted for in the calculations. A product superposition technique is employed at the source which allows a more efficient usage of degrees of freedom when the relay and the destination have different training requirements. Varying configurations of coherence times are studied. The interesting case where the different link coherence intervals are not a multiple of each other, and therefore the coherence intervals do not align, is studied. Relay scheduling is combined with the product superposition to obtain further gains in degrees of freedom. The impact of coherence diversity is further studied in the presencemore »of multiple parallel relays.« less
    Free, publicly-accessible full text available January 1, 2023