Mixture of Hidden-Dimensions Transformer

Transformer models encounter challenges in scaling hidden dimensions efficiently, as uniformly increasing them inflates computational and memory costs while failing to emphasize the most relevant features for each token. For further understanding, we study hidden dimension sparsity and observe that trained Transformers utilize only a small fraction of token dimensions, revealing an "activation flow" pattern. Notably, there are shared sub-dimensions with sustained activation across multiple consecutive tokens and specialized sub-dimensions uniquely activated for each token. To better model token-relevant sub-dimensions, we propose MoHD (Mixture of Hidden Dimensions), a sparse conditional activation architecture. Particularly, MoHD employs shared sub-dimensions for common token features and a routing mechanism to dynamically activate specialized sub-dimensions. To mitigate potential information loss from sparsity, we design activation scaling and group fusion mechanisms to preserve activation flow. In this way, MoHD expands hidden dimensions with negligible increases in computation or parameters, efficient training and inference while maintaining performance. Evaluations across 10 NLP tasks show that MoHD surpasses Vanilla Transformers in parameter efficiency and task performance. It achieves 1.7% higher performance with 50% fewer activation parameters and 3.7% higher performance with a 3x parameter expansion at constant activation cost. MOHD offers a new perspective for scaling the model, showcasing the potential of hidden dimension sparsity to boost efficiency

MONOPOLY: Learning to Price Public Facilities for Revaluing Private Properties with Large-Scale Urban Data

The value assessment of private properties is an attractive but challenging task which is widely concerned by a majority of people around the world. A prolonged topic among us is ``\textit{how much is my house worth?}''. To answer this question, most experienced agencies would like to price a property given the factors of its attributes as well as the demographics and the public facilities around it. However, no one knows the exact prices of these factors, especially the values of public facilities which may help assess private properties. In this paper, we introduce our newly launched project ``Monopoly'' (named after a classic board game) in which we propose a distributed approach for revaluing private properties by learning to price public facilities (such as hospitals etc.) with the large-scale urban data we have accumulated via Baidu Maps. To be specific, our method organizes many points of interest (POIs) into an undirected weighted graph and formulates multiple factors including the virtual prices of surrounding public facilities as adaptive variables to parallelly estimate the housing prices we know. Then the prices of both public facilities and private properties can be iteratively updated according to the loss of prediction until convergence. We have conducted extensive experiments with the large-scale urban data of several metropolises in China. Results show that our approach outperforms several mainstream methods with significant margins. Further insights from more in-depth discussions demonstrate that the ``Monopoly'' is an innovative application in the interdisciplinary field of business intelligence and urban computing, and it will be beneficial to tens of millions of our users for investments and to the governments for urban planning as well as taxation.

DuMapper: Towards Automatic Verification of Large-Scale POIs with Street Views at Baidu Maps

With the increased popularity of mobile devices, Web mapping services have become an indispensable tool in our daily lives. To provide user-satisfied services, such as location searches, the point of interest (POI) database is the fundamental infrastructure, as it archives multimodal information on billions of geographic locations closely related to people's lives, such as a shop or a bank. Therefore, verifying the correctness of a large-scale POI database is vital. To achieve this goal, many industrial companies adopt volunteered geographic information (VGI) platforms that enable thousands of crowdworkers and expert mappers to verify POIs seamlessly; but to do so, they have to spend millions of dollars every year. To save the tremendous labor costs, we devised DuMapper, an automatic system for large-scale POI verification with the multimodal street-view data at Baidu Maps. DuMapper takes the signboard image and the coordinates of a real-world place as input to generate a low-dimensional vector, which can be leveraged by ANN algorithms to conduct a more accurate search through billions of archived POIs in the database for verification within milliseconds. It can significantly increase the throughput of POI verification by $50$ times. DuMapper has already been deployed in production since \DuMPOnline, which dramatically improves the productivity and efficiency of POI verification at Baidu Maps. As of December 31, 2021, it has enacted over $405$ million iterations of POI verification within a 3.5-year period, representing an approximate workload of $800$ high-performance expert mappers.

Quantum Neural Network for Accelerated Magnetic Resonance Imaging

Magnetic resonance image reconstruction starting from undersampled k-space data requires the recovery of many potential nonlinear features, which is very difficult for algorithms to recover these features. In recent years, the development of quantum computing has discovered that quantum convolution can improve network accuracy, possibly due to potential quantum advantages. This article proposes a hybrid neural network containing quantum and classical networks for fast magnetic resonance imaging, and conducts experiments on a quantum computer simulation system. The experimental results indicate that the hybrid network has achieved excellent reconstruction results, and also confirm the feasibility of applying hybrid quantum-classical neural networks into the image reconstruction of rapid magnetic resonance imaging.

Optimized Magnetic Resonance Fingerprinting Using Ziv-Zakai Bound

Magnetic Resonance Fingerprinting (MRF) has emerged as a promising quantitative imaging technique within the field of Magnetic Resonance Imaging (MRI), offers comprehensive insights into tissue properties by simultaneously acquiring multiple tissue parameter maps in a single acquisition. Sequence optimization is crucial for improving the accuracy and efficiency of MRF. In this work, a novel framework for MRF sequence optimization is proposed based on the Ziv-Zakai bound (ZZB). Unlike the Cram\'er-Rao bound (CRB), which aims to enhance the quality of a single fingerprint signal with deterministic parameters, ZZB provides insights into evaluating the minimum mismatch probability for pairs of fingerprint signals within the specified parameter range in MRF. Specifically, the explicit ZZB is derived to establish a lower bound for the discrimination error in the fingerprint signal matching process within MRF. This bound illuminates the intrinsic limitations of MRF sequences, thereby fostering a deeper understanding of existing sequence performance. Subsequently, an optimal experiment design problem based on ZZB was formulated to ascertain the optimal scheme of acquisition parameters, maximizing discrimination power of MRF between different tissue types. Preliminary numerical experiments show that the optimized ZZB scheme outperforms both the conventional and CRB schemes in terms of the reconstruction accuracy of multiple parameter maps.

FlashMask: Efficient and Rich Mask Extension of FlashAttention

The computational and memory demands of vanilla attention scale quadratically with the sequence length $N$, posing significant challenges for processing long sequences in Transformer models. FlashAttention alleviates these challenges by eliminating the $O(N^2)$ memory dependency and reducing attention latency through IO-aware memory optimizations. However, its native support for certain attention mask types is limited, and it does not inherently accommodate more complex masking requirements. Previous approaches resort to using dense masks with $O(N^2)$ memory complexity, leading to inefficiencies. In this paper, we propose FlashMask, an extension of FlashAttention that introduces a column-wise sparse representation of attention masks. This approach efficiently represents a wide range of mask types and facilitates the development of optimized kernel implementations. By adopting this novel representation, FlashMask achieves linear memory complexity $O(N)$, suitable for modeling long-context sequences. Moreover, this representation enables kernel optimizations that eliminate unnecessary computations by leveraging sparsity in the attention mask, without sacrificing computational accuracy, resulting in higher computational efficiency. We evaluate FlashMask's performance in fine-tuning and alignment training of LLMs such as SFT, LoRA, DPO, and RM. FlashMask achieves significant throughput improvements, with end-to-end speedups ranging from 1.65x to 3.22x compared to existing FlashAttention dense method. Additionally, our kernel-level comparisons demonstrate that FlashMask surpasses the latest counterpart, FlexAttention, by 12.1% to 60.7% in terms of kernel TFLOPs/s, achieving 37.8% to 62.3% of the theoretical maximum FLOPs/s on the A100 GPU. The code is open-sourced on PaddlePaddle and integrated into PaddleNLP, supporting models with over 100 billion parameters for contexts up to 128K tokens.

MoistNet: Machine Vision-based Deep Learning Models for Wood Chip Moisture Content Measurement

Quick and reliable measurement of wood chip moisture content is an everlasting problem for numerous forest-reliant industries such as biofuel, pulp and paper, and bio-refineries. Moisture content is a critical attribute of wood chips due to its direct relationship with the final product quality. Conventional techniques for determining moisture content, such as oven-drying, possess some drawbacks in terms of their time-consuming nature, potential sample damage, and lack of real-time feasibility. Furthermore, alternative techniques, including NIR spectroscopy, electrical capacitance, X-rays, and microwaves, have demonstrated potential; nevertheless, they are still constrained by issues related to portability, precision, and the expense of the required equipment. Hence, there is a need for a moisture content determination method that is instant, portable, non-destructive, inexpensive, and precise. This study explores the use of deep learning and machine vision to predict moisture content classes from RGB images of wood chips. A large-scale image dataset comprising 1,600 RGB images of wood chips has been collected and annotated with ground truth labels, utilizing the results of the oven-drying technique. Two high-performing neural networks, MoistNetLite and MoistNetMax, have been developed leveraging Neural Architecture Search (NAS) and hyperparameter optimization. The developed models are evaluated and compared with state-of-the-art deep learning models. Results demonstrate that MoistNetLite achieves 87% accuracy with minimal computational overhead, while MoistNetMax exhibits exceptional precision with a 91% accuracy in wood chip moisture content class prediction. With improved accuracy and faster prediction speed, our proposed MoistNet models hold great promise for the wood chip processing industry.

Activation Function Optimization Scheme for Image Classification

Activation function has a significant impact on the dynamics, convergence, and performance of deep neural networks. The search for a consistent and high-performing activation function has always been a pursuit during deep learning model development. Existing state-of-the-art activation functions are manually designed with human expertise except for Swish. Swish was developed using a reinforcement learning-based search strategy. In this study, we propose an evolutionary approach for optimizing activation functions specifically for image classification tasks, aiming to discover functions that outperform current state-of-the-art options. Through this optimization framework, we obtain a series of high-performing activation functions denoted as Exponential Error Linear Unit (EELU). The developed activation functions are evaluated for image classification tasks from two perspectives: (1) five state-of-the-art neural network architectures, such as ResNet50, AlexNet, VGG16, MobileNet, and Compact Convolutional Transformer which cover computationally heavy to light neural networks, and (2) eight standard datasets, including CIFAR10, Imagenette, MNIST, Fashion MNIST, Beans, Colorectal Histology, CottonWeedID15, and TinyImageNet which cover from typical machine vision benchmark, agricultural image applications to medical image applications. Finally, we statistically investigate the generalization of the resultant activation functions developed through the optimization scheme. With a Friedman test, we conclude that the optimization scheme is able to generate activation functions that outperform the existing standard ones in 92.8% cases among 28 different cases studied, and $-x\cdot erf(e^{-x})$ is found to be the best activation function for image classification generated by the optimization scheme.

HGAMN: Heterogeneous Graph Attention Matching Network for Multilingual POI Retrieval at Baidu Maps

The increasing interest in international travel has raised the demand of retrieving point of interests in multiple languages. This is even superior to find local venues such as restaurants and scenic spots in unfamiliar languages when traveling abroad. Multilingual POI retrieval, enabling users to find desired POIs in a demanded language using queries in numerous languages, has become an indispensable feature of today's global map applications such as Baidu Maps. This task is non-trivial because of two key challenges: (1) visiting sparsity and (2) multilingual query-POI matching. To this end, we propose a Heterogeneous Graph Attention Matching Network (HGAMN) to concurrently address both challenges. Specifically, we construct a heterogeneous graph that contains two types of nodes: POI node and query node using the search logs of Baidu Maps. To alleviate challenge \#1, we construct edges between different POI nodes to link the low-frequency POIs with the high-frequency ones, which enables the transfer of knowledge from the latter to the former. To mitigate challenge \#2, we construct edges between POI and query nodes based on the co-occurrences between queries and POIs, where queries in different languages and formulations can be aggregated for individual POIs. Moreover, we develop an attention-based network to jointly learn node representations of the heterogeneous graph and further design a cross-attention module to fuse the representations of both types of nodes for query-POI relevance scoring. Extensive experiments conducted on large-scale real-world datasets from Baidu Maps demonstrate the superiority and effectiveness of HGAMN. In addition, HGAMN has already been deployed in production at Baidu Maps, and it successfully keeps serving hundreds of millions of requests every day.

MR Optimized Reconstruction of Simultaneous Multi-Slice Imaging Using Diffusion Model

Diffusion model has been successfully applied to MRI reconstruction, including single and multi-coil acquisition of MRI data. Simultaneous multi-slice imaging (SMS), as a method for accelerating MR acquisition, can significantly reduce scanning time, but further optimization of reconstruction results is still possible. In order to optimize the reconstruction of SMS, we proposed a method to use diffusion model based on slice-GRAPPA and SPIRiT method. approach: Specifically, our method characterizes the prior distribution of SMS data by score matching and characterizes the k-space redundant prior between coils and slices based on self-consistency. With the utilization of diffusion model, we achieved better reconstruction results.The application of diffusion model can further reduce the scanning time of MRI without compromising image quality, making it more advantageous for clinical application