Search

We study the continuous vector-valued Witsenhausen counterexample through the lens of empirical coordination coding. We characterize the region of achievable pairs of costs in three scenarios: (i) causal encoding and causal decoding, (ii) causal encoding and causal decoding with channel feedback, and (iii) causal encoding and noncausal decoding with channel feedback. In these vector-valued versions of the problem, the optimal coding schemes must rely on a time-sharing strategy, since the region of achievable pairs of costs might not be convex in the scalar version of the problem. We examine the role of the channel feedback when the encoder is causal and the decoder is either causal or non-causal, and we show that feedback improves the performance, only when the decoder is non-causal., Comment: Accepted by IEEE Information Theory Workshop (ITW) 2024

In this study, we investigate a vector-valued Witsenhausen model where the second decision maker (DM) acquires a vector of observations before selecting a vector of estimations. Here, the first DM acts causally whereas the second DM estimates non-causally. When the vector length grows, we characterize, via a single-letter expression, the optimal trade-off between the power cost at the first DM and the estimation cost at the second DM. In this paper, we show that the best linear scheme is achieved by using the time-sharing method between two affine strategies, which coincides with the convex envelope of the solution of Witsenhausen in 1968. Here also, Witsenhausen's two-point strategy and the scheme of Grover and Sahai in 2010 where both devices operate non-causally, outperform our best linear scheme. Therefore, gains obtained with block-coding schemes are only attainable if all DMs operate non-causally.

We investigate the Witsenhausen counterexample in a continuous vector-valued context with a causal encoder and noncausal decoder. Our main result is the optimal single-letter condition that characterizes the set of achievable Witsenhausen power costs and estimation costs, leveraging a modified weak typicality approach. In particular, we accommodate our power analysis to the causal encoder constraint, and provide an improved distortion error analysis for the challenging estimation of the interim state. Interestingly, the idea of dual role of control is explicitly captured by the two auxiliary random variables., Comment: This work has been submitted to 2024 IEEE International Symposium on Information Theory (ISIT 2024)

Achieving high-resolution Direction of Arrival (DoA) recovery typically requires high Signal to Noise Ratio (SNR) and a sufficiently large number of snapshots. This paper presents NUV-DoA algorithm, that augments Bayesian sparse reconstruction with spatial filtering for super-resolution DoA estimation. By modeling each direction on the azimuth's grid with the sparsity-promoting normal with unknown variance (NUV) prior, the non-convex optimization problem is reduced to iteratively reweighted least-squares under Gaussian distribution, where the mean of the snapshots is a sufficient statistic. This approach not only simplifies our solution but also accurately detects the DoAs. We utilize a hierarchical approach for interference cancellation in multi-source scenarios. Empirical evaluations show the superiority of NUV-DoA, especially in low SNRs, compared to alternative DoA estimators., Comment: This paper has 5 pages including reference, 11 figures. This paper has been accepted to ICASSP 2024 - 2024 International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

The recent emergence of joint CTC-Attention model shows significant improvement in automatic speech recognition (ASR). The improvement largely lies in the modeling of linguistic information by decoder. The decoder joint-optimized with an acoustic encoder renders the language model from ground-truth sequences in an auto-regressive manner during training. However, the training corpus of the decoder is limited to the speech transcriptions, which is far less than the corpus needed to train an acceptable language model. This leads to poor robustness of decoder. To alleviate this problem, we propose linguistic-enhanced transformer, which introduces refined CTC information to decoder during training process, so that the decoder can be more robust. Our experiments on AISHELL-1 speech corpus show that the character error rate (CER) is relatively reduced by up to 7%. We also find that in joint CTC-Attention ASR model, decoder is more sensitive to linguistic information than acoustic information., Comment: Accepted by ECAISS2022, The Fourth International Workshop on Edge Computing and Artificial Intelligence based Sensor-Cloud System

In the context of simulating precision laser interferometers, we compare via several examples two wavefront decomposition methods: the Mode Expansion Method (MEM) and the Gaussian beam decomposition (GBD) for their precision and applicability. To judge the performance of these methods, we define different types of errors and study their properties. We specify how the two methods can be fairly compared and based on that, the quality of the MEM and GBD are compared in several examples. We test here cases for which analytic results are available, i.e., non-clipped circular and general astigmatic Gaussian beams, as well as clipped circular Gaussian beams, in the near-, far-, and extreme far-field of millions of kilometers occurring in space-gravitational wave detectors. Additionally, we compare the methods for aberrated wavefronts and the interaction with optical components by testing reflections from differently curved mirrors. We find that both methods can be generally used for decomposing non-Gaussian beams. However, which method is more accurate depends on the optical system and simulation settings. In the given examples, the MEM more accurately describes non-clipped Gaussian beams, while for clipped Gaussian beams and the interaction with surfaces, the GBD is more precise.

With the development of Internet technology, the phenomenon of information overload is becoming more and more obvious. It takes a lot of time for users to obtain the information they need. However, keyphrases that summarize document information highly are helpful for users to quickly obtain and understand documents. For academic resources, most existing studies extract keyphrases through the title and abstract of papers. We find that title information in references also contains author-assigned keyphrases. Therefore, this article uses reference information and applies two typical methods of unsupervised extraction methods (TF*IDF and TextRank), two representative traditional supervised learning algorithms (Na\"ive Bayes and Conditional Random Field) and a supervised deep learning model (BiLSTM-CRF), to analyze the specific performance of reference information on keyphrase extraction. It is expected to improve the quality of keyphrase recognition from the perspective of expanding the source text. The experimental results show that reference information can increase precision, recall, and F1 of automatic keyphrase extraction to a certain extent. This indicates the usefulness of reference information on keyphrase extraction of academic papers and provides a new idea for the following research on automatic keyphrase extraction.

The traditional recommendation systems mainly use offline user data to train offline models, and then recommend items for online users, thus suffering from the unreliable estimation of user preferences based on sparse and noisy historical data. Conversational Recommendation System (CRS) uses the interactive form of the dialogue systems to solve the intrinsic problems of traditional recommendation systems. However, due to the lack of contextual information modeling, the existing CRS models are unable to deal with the exploitation and exploration (E&E) problem well, resulting in the heavy burden on users. To address the aforementioned issue, this work proposes a contextual information enhancement model tailored for CRS, called Knowledge Graph-enhanced Sampling (KGenSam). KGenSam integrates the dynamic graph of user interaction data with the external knowledge into one heterogeneous Knowledge Graph (KG) as the contextual information environment. Then, two samplers are designed to enhance knowledge by sampling fuzzy samples with high uncertainty for obtaining user preferences and reliable negative samples for updating recommender to achieve efficient acquisition of user preferences and model updating, and thus provide a powerful solution for CRS to deal with E&E problem. Experimental results on two real-world datasets demonstrate the superiority of KGenSam with significant improvements over state-of-the-art methods.