Search

In this paper, we present Duplex Conversation, a multi-turn, multimodal spoken dialogue system that enables telephone-based agents to interact with customers like a human. We use the concept of full-duplex in telecommunication to demonstrate what a human-like interactive experience should be and how to achieve smooth turn-taking through three subtasks: user state detection, backchannel selection, and barge-in detection. Besides, we propose semi-supervised learning with multimodal data augmentation to leverage unlabeled data to increase model generalization. Experimental results on three sub-tasks show that the proposed method achieves consistent improvements compared with baselines. We deploy the Duplex Conversation to Alibaba intelligent customer service and share lessons learned in production. Online A/B experiments show that the proposed system can significantly reduce response latency by 50%., Comment: Accepted by KDD 2022, ADS track

Polar code visibly approaches channel capacity in practice and is thereby a constituent code of the 5G standard. Compared to low-density parity-check code, however, the performance of short-length polar code has rooms for improvement that could hinder its adoption by a wider class of applications. As part of the program that addresses the performance issue at short length, it is crucial to understand how fast binary memoryless symmetric channels polarize. A number, called scaling exponent, was defined to measure the speed of polarization and several estimates of the scaling exponent were given in literature. As of 2022, the tightest overestimate is 4.714 made by Mondelli, Hassani, and Urbanke in 2015. We lower the overestimate to 4.63., Comment: 15 pages, 13 figures, 1 table

We construct an explicit family of 3-XOR instances hard for $\Omega(n)$-levels of the Sum-of-Squares (SoS) semi-definite programming hierarchy. Not only is this the first explicit construction to beat brute force search (beyond low-order improvements (Tulsiani 2021, Pratt 2021)), combined with standard gap amplification techniques it also matches the (optimal) hardness of random instances up to imperfect completeness (Grigoriev TCS 2001, Schoenebeck FOCS 2008). Our result is based on a new form of small-set high dimensional expansion (SS-HDX) inspired by recent breakthroughs in locally testable and quantum LDPC codes. Adapting the recent framework of Dinur, Filmus, Harsha, and Tulsiani (ITCS 2021) for SoS lower bounds from the Ramanujan complex to this setting, we show any (bounded-degree) SS-HDX can be transformed into a highly unsatisfiable 3-XOR instance that cannot be refuted by $\Omega(n)$-levels of SoS. We then show Leverrier and Z\'emor's (Arxiv 2022) recent qLDPC construction gives the desired explicit family of bounded-degree SS-HDX. Incidentally, this gives the strongest known form of bi-directional high dimensional expansion to date.

With the good performance of deep learning algorithms in the field of computer vision (CV), the convolutional neural network (CNN) architecture has become a main backbone of the computer vision task. With the widespread use of mobile devices, neural network models based on platforms with low computing power are gradually being paid attention. However, due to the limitation of computing power, deep learning algorithms are usually not available on mobile devices. This paper proposes a lightweight convolutional neural network, TripleNet, which can operate easily on Raspberry Pi. Adopted from the concept of block connections in ThreshNet, the newly proposed network model compresses and accelerates the network model, reduces the amount of parameters of the network, and shortens the inference time of each image while ensuring the accuracy. Our proposed TripleNet and other state-of-the-art (SOTA) neural networks perform image classification experiments with the CIFAR-10 and SVHN datasets on Raspberry Pi. The experimental results show that, compared with GhostNet, MobileNet, ThreshNet, EfficientNet, and HarDNet, the inference time of TripleNet per image is shortened by 15%, 16%, 17%, 24%, and 30%, respectively. The detail codes of this work are available at https://github.com/RuiyangJu/TripleNet.

Quantum low-density parity-check (qLDPC) codes are quantum stabilizer codes where each stabilizer acts on a constant number of qubits and each qubit is acted on by a constant number of stabilizers. We study qLDPC codes constructed from balanced products and lossless expanders. We found that assuming the existence of 2-sided lossless expander graphs with free group action, the resulting qLDPC codes have constant rate, linear distance, and linear time decoders., Comment: arXiv admin note: text overlap with arXiv:2201.11369

With the achievements of Transformer in the field of natural language processing, the encoder-decoder and the attention mechanism in Transformer have been applied to computer vision. Recently, in multiple tasks of computer vision (image classification, object detection, semantic segmentation, etc.), state-of-the-art convolutional neural networks have introduced some concepts of Transformer. This proves that Transformer has a good prospect in the field of image recognition. After Vision Transformer was proposed, more and more works began to use self-attention to completely replace the convolutional layer. This work is based on Vision Transformer, combined with the pyramid architecture, using Split-transform-merge to propose the group encoder and name the network architecture Aggregated Pyramid Vision Transformer (APVT). We perform image classification tasks on the CIFAR-10 dataset and object detection tasks on the COCO 2017 dataset. Compared with other network architectures that use Transformer as the backbone, APVT has excellent results while reducing the computational cost. We hope this improved strategy can provide a reference for future Transformer research in computer vision., Comment: 6 pages, 5 figures

Understanding the internal spin structure of the nucleon still remains a challenge in strong interaction physics. Transversity, which describes the transverse spin structure of quarks in a transversely polarized proton, is poorly constrained by experimental data. Since it is chiral-odd, it can only be accessed through channels that couple to other chiral-odd distributions, like the Collins fragmentation functions (so-called Collins effect) or the interference fragmentation functions. Recently, a detailed calculation using the soft-collinear effective theory found that the Collins effect in $pp$ collisions involves a mixture of collinear and transverse momentum dependent (TMD) factorization. The Collins effect provides a direct probe to the Collins fragmentation function and enables testing of its evolution, universality and factorization breaking in the transverse momentum dependent formalism. In 2018, STAR published the first measurements of Collins asymmetries for charged pions in jets in polarized $pp$ collisions at $\sqrt{s}$ = 500 GeV based on data taken during 2011. These measurements probe $Q^2$ scales one to two orders of magnitude larger than similar measurements in semi-inclusive deep-inelastic scattering (SIDIS) and the results are consistent with predictions based on global analyses of $e^{+}e^{-}$ and SIDIS data. In 2012 and 2015, STAR collected $\sim$14 $\mathrm{pb^{-1}}$ and $\sim$52 $\mathrm{pb^{-1}}$ of transversely polarized $pp$ data at $\sqrt{s}$ = 200 GeV, respectively. These datasets provide the most precise measurement of the Collins effect in 200 GeV $pp$ collisions to date, especially at the quark momentum fractions $0.1 \le x \le 0.4$. These proceedings present the preliminary results for Collins asymmetries of identified pions in jets in $pp$ collisions at $\sqrt{s}$ = 200 GeV and comparisons to theory predictions., Comment: Contribution to the Proceedings of the 24th International Spin Symposium (SPIN2021)

A sentence may express sentiments on multiple aspects. When these aspects are associated with different sentiment polarities, a model's accuracy is often adversely affected. We observe that multiple aspects in such hard sentences are mostly expressed through multiple clauses, or formally known as elementary discourse units (EDUs), and one EDU tends to express a single aspect with unitary sentiment towards that aspect. In this paper, we propose to consider EDU boundaries in sentence modeling, with attentions at both word and EDU levels. Specifically, we highlight sentiment-bearing words in EDU through word-level sparse attention. Then at EDU level, we force the model to attend to the right EDU for the right aspect, by using EDU-level sparse attention and orthogonal regularization. Experiments on three benchmark datasets show that our simple EDU-Attention model outperforms state-of-the-art baselines. Because EDU can be automatically segmented with high accuracy, our model can be applied to sentences directly without the need of manual EDU boundary annotation., Comment: Accepted in PAKDD2022

A locally testable code (LTC) is an error correcting code with a property tester. The tester tests if a word is codeword by reading constant random bits and rejects the word with probability proportional to the distance from the word to the closest codeword. An important open question until recently is whether there exist $c^3$-LTCs which are LTCs with constant rate, constant relative distance and constant locality. In this work, we construct a new LTC family using 1-sided lossless expanders and balanced products.

With the continuous development of neural networks for computer vision tasks, more and more network architectures have achieved outstanding success. As one of the most advanced neural network architectures, DenseNet shortcuts all feature maps to solve the model depth problem. Although this network architecture has excellent accuracy with low parameters, it requires an excessive inference time. To solve this problem, HarDNet reduces the connections between the feature maps, making the remaining connections resemble harmonic waves. However, this compression method may result in a decrease in the model accuracy and an increase in the parameters and model size. This network architecture may reduce the memory access time, but its overall performance can still be improved. Therefore, we propose a new network architecture, ThreshNet, using a threshold mechanism to further optimize the connection method. Different numbers of connections for different convolution layers are discarded to accelerate the inference of the network. The proposed network has been evaluated with image classification using CIFAR 10 and SVHN datasets under platforms of NVIDIA RTX 3050 and Raspberry Pi 4. The experimental results show that, compared with HarDNet68, GhostNet, MobileNetV2, ShuffleNet, and EfficientNet, the inference time of the proposed ThreshNet79 is 5%, 9%, 10%, 18%, and 20% faster, respectively. The number of parameters of ThreshNet95 is 55% less than that of HarDNet85. The new model compression and model acceleration methods can speed up the inference time, enabling network models to operate on mobile devices., Comment: IEEE Access

CholeskyQR is a simple and fast QR decomposition via Cholesky decomposition, while it has been considered highly sensitive to the condition number. In this paper, we provide a randomized preconditioner framework for CholeskyQR algorithm. Under this framework, two methods (randomized LU-CholeskyQR and randomized QR-CholeskyQR) are proposed and discussed. We prove the proposed preconditioners can effectively reduce the condition number, which is also demonstrated by numerical tests. Abundant numerical tests indicate our methods are more stable and faster than all the existing algorithms and have good scalability., Comment: 23 pages, 11 figures, 6 tables