Adversarial for Sequential Recommendation Walking in the Multi-Latent Space

Ziyi Zhu, Zeyu Wang, Zhizhong Wu, Yiqian Zhang, Shi Bo — Wed, 24 Jul 2024 00:00:00 +0530

Recently, sequential recommendation plays a critical role in our daily life, since it serves as personalized information filters to dis- cover popular users’ preferred products over time. Due to the success of the adversarial learning, a mass of research efforts start to strengthen sequential recommendation by the adversarial learning, which is able to learn complex underlying data distribution.

However, existing adversarial sequential recommendation methods suffer from mode collapse and unexplained prediction. To boost the diversity, performance, and interpretability of sequential recommendation system, we propose a novel temporal-aware adversarial framework, namely TSRGAN.

In principle, the input of traditional adversarial-based recommendation system is a noise variable sampled from normal distribution. We argue that it is hard to generate an item cover complex users’ preferences(e.g. price, brand and item style) using a single latent space. Therefore, our model employs multiple latent space to generate plausible item which matches user’ preferences from multiple views(e.g. Movie style, Movie release date).

Besides, previous adversarial-based recommenders focus on generating active item, but they omits that user’s favour is not in- variable. With GANs terminology, the recommenders only will be rewarded when seeking the peak mode, but it neglects minor mode, in other word mode collapse. In order to alleviate this issue, we design a novel diversity reward function and diversify regularization to encourage the model exploring minor mode over time and guarantee generating diversity item with reasonable.

Concretely, we propose multiple learnable latent codes to generate item matching user’s preferences from different views, then we leverage the diversity reward signal to shape the distribution of multiple latent space over time. It means that the multiple latent space are sampled form different distribution instead of Gaussian distribution. Such a manipulation of the latent space can be treated as walking from plain distribution latent space to diversity distributions latent space. Further, the reward signal is modified over time, therefore, our methods names "Temporal-aware" adversarial framework.

In short, our model has two sequential stages: encode the user’ characteristics and historical behaviours under multiple latent space with the Self Attention-based generator(G), and discriminator(D) try to distinguish the generator’s output item from the ground ruth. Besides, discriminator attempt to apply reward signal to shape the latent space distribution time by time. Extensive experiments demonstrate remarkable performance with interpretability improvement against the state-of-the-art baselines.

LLM Connection Graphs for Global Feature Extraction in Point Cloud Analysis

Zeyu Wang, Yue Zhu, Minghao Chen, Minghao Liu, Weijian Qin — Tue, 30 Jul 2024 00:00:00 +0530

Graph convolutional networks (GCNs) have effectively utilized local connections for point cloud analysis. How- ever, capturing distant dependencies (i.e., global features) with a single local connection graph, such as the Euclidean k-nearest neighbor graph, remains challenging. To ad- dress this, we introduce the Multi-Space Graph Convolutional Network (PointGCNN), which leverages reinforcement learning to adaptively construct connection graphs in multiple latent spaces, integrating both local and non-local dependencies. Initially, we encode and concatenate low- level local features from Euclidean and Eigenvalue spaces. Convolution layers are then hierarchically built, with each layer forming dynamic connection graphs to guide the propagation of low-level features. [1,2,3,4,11,14,16]These implicitly constructed graphs enable our model to uncover hidden dependencies. The assorted connections from different graphs support the extraction of fine-grained features from various perspectives, enhancing complex scene recognition. Thus, our model can capture multiple global contexts beyond the local scope of a single space, providing strong robustness against perturbations. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on two major public point cloud benchmarks.

Applied Science and Biotechnology Journal for Advanced Research

Adversarial for Sequential Recommendation Walking in the Multi-Latent Space

LLM Connection Graphs for Global Feature Extraction in Point Cloud Analysis