23 accepted papers
Managing agent thought and observation during multi-turn agent-environment interactions is an emerging strategy to improve agent efficiency. However, existing studies treat the entire interaction trajectories equally, overlooking the thought necessity and observation utility varies across turns. To …
Multimodal Large Language Model (MLLM) agents facilitate Graphical User Interface (GUI) automation but struggle with long-horizon, cross-application tasks due to limited context windows. While memory systems provide a viable solution, existing paradigms struggle to adapt to dynamic GUI environments,…
Representation learning is fundamental to modern machine learning, powering applications such as text retrieval and multimodal understanding. However, learning robust and generalizable representations remains challenging. While prior work has demonstrated that active noise injection, a form of data
Recent advances have empowered large language models (LLMs) with remarkable fine-grained instruction-following capabilities in text generation tasks. However, embedding methods typically rely solely on the hidden state of the input's last token, limiting their ability to capture complete semantic si…
Recent generative data augmentation methods conditioned on both image and text prompts struggle to balance between fidelity and diversity, as it is challenging to preserve essential image details while aligning with varied text prompts. This challenge arises because representations in the synthesis…
In this work, we consider a multi-server federated learning (FL) framework, referred to as Confederated Learning (CFL), in order to accommodate a larger number of users. To reduce the communication overhead of the CFL system, we propose a linearly convergent stochastic gradient method. The proposed…
We consider the problem of maximizing the minimum user achievable rate for downlink massive MIMO multi-user systems. The max-min formulation promotes fairness such that all users enjoy a similar quality of wireless service. Nevertheless, solving the max-min beamforming problem is challenging due to…
Existing losses used in deep metric learning (DML) for image retrieval often lead to highly non-uniform intra-class and inter-class representation structures across test classes and data distributions. When combined with the common practice of using a fixed threshold to declare a match, this gives r…
In this paper, we provide an in-depth study of Stochastic Backpropagation (SBP) when training deep neural networks for standard image classification and object detection tasks. During backward propagation, SBP calculates gradients by using only a subset of feature maps to save GPU memory and computa…
Aspect-based sentiment analysis (ABSA) has received increasing attention recently. ABSA can be divided into multiple tasks according to the different extracted elements. Existing generative methods usually treat the output as a whole string rather than the combination of different elements and only…
"With the shift towards on-device deep learning, ensuring a consistent behavior of an AI service across diverse compute platforms becomes tremendously important. Our work tackles the emergent problem of reducing predictive in-consistencies arising as negative flips: test samples that are correctly p…
We consider the problem of joint wideband spectrum sensing and carrier frequency estimation in a sub-Nyquist sampling framework. Specifically, a multi-antenna receiver is used to estimate the carrier frequencies and power spectra of multiple narrowband transmissions that spread over a wide frequency…
Quantization plays an important role in the energy-efficient deployment of Deep Neural Networks (DNNs) on resource-limited devices. Post-training quantization is highly desirable since it does not require retraining or access to the full training dataset. The well-established uniform scheme for post…
This paper concerns wideband direction of arrival (DoA) estimation with sparse linear arrays (SLAs). We rely on the assumption that the power spectrum of the wideband sources is the same up to a scaling factor, which could in theory allow us to resolve not only more sources than the number of antenn…
The problem of beam alignment for millimeter wave (mm-Wave) communications is studied in this paper. We show that, by exploiting the sparse scattering nature of mmWave channels, the beam alignment problem can be formulated as a sparse encoding and phaseless decoding problem, which involves finding a…
We propose DGG: D eep clustering via a G aussian-mixture variational autoencoder (VAE) with G raph embedding. To facilitate clustering, we apply Gaussian mixture model (GMM) as the prior in VAE. To handle data with complex spread, we apply graph embedding. Our idea is that graph information which…
This paper considers a two-receiver multiple-input multiple-output (MIMO) Gaussian broadcast channel model with integrated services. Specifically, two sorts of service messages are combined and served simultaneously: one multicast message intended for both receivers and one confidential message inte…
This paper concerns detecting the frequency components from a spectral sparse, undersampled signal. This problem is also called super-resolution line spectral estimation because the frequencies can take arbitrary continuous values. The prior knowledge of the frequency distribution is often available…
In this paper, a new iteratively reweighted least squares method is proposed for recovery of block-sparse signals with unknown cluster patterns. In many practical applications, sparse signals have block-sparse structures with nonzero coefficients occurring in clusters, while the prior information of…
This paper considers adaptive signal detection in stochastic homogeneous environments where the disturbance covariance matrix of both test and training signals, R, is assumed to be a random matrix with a priori knowledge of R. Unlike existing detectors assuming a known hyperparameter associated with…
This paper considers secrecy communication from a signal processing point of view, and studies the maximal achievable secrecy degrees of freedoms (S.D.o.F.) of a helper-assisted Gaussian wiretap channel, consisting of a source, a legitimate receiver, an eavesdropper and an external helper. Each term…
We consider a dictionary learning problem aimed at designing a dictionary such that the signals admits a sparse or an approximate sparse representation over the learned dictionary. The problem finds a variety of applications including image denoising, feature extraction, etc. In this paper, we propo…
In this paper, we study the problem of sparse signal recovery when partial but partly erroneous prior knowledge of the signal's support is available. Based on the conventional sparse Bayesian learning framework, we propose an improved hierarchical prior model. The proposed modeling constitutes a thr…