Sequential Recommendation for Cold-start Users with Meta Transitional Learning

A fundamental challenge for sequential recommenders is to capture the sequential patterns of users toward modeling how users transit among items. In many practical scenarios, however, there are a great number of cold-start users with only minimal logged interactions. As a result, existing sequential recommendation models will lose their predictive power due to the difficulties in learning sequential patterns over users with only limited interactions. In this work, we aim to improve sequential recommendation for cold-start users with a novel framework named MetaTL, which learns to model the transition patterns of users through meta-learning. Specifically, the proposed MetaTL: (i) formulates sequential recommendation for cold-start users as a few-shot learning problem; (ii) extracts the dynamic transition patterns among users with a translation-based architecture; and (iii) adopts meta transitional learning to enable fast learning for cold-start users with only limited interactions, leading to accurate inference of sequential interactions.

Which Milky Way masses are consistent with the slightly declining 5-25 kpc rotation curve?

Discoveries of extended rotation curves have suggested the presence of dark matter in spiral galaxy haloes. It has led to many studies that estimated the galaxy total mass, mostly by using the Navarro Frenk and White (NFW) density profile. We aim at verifying how the choice of the dark-matter profile may affect the predicted values of extrapolated total masses. We have considered the recent Milky Way (MW) rotation curve, firstly because of its unprecedented accuracy, and secondly because the Galactic disk is amongst the least affected by past major mergers having fully reshaped the initial disk. We find that the use of NFW profile (or its generalized form, gNFW) for calculating the dark-matter contribution to the MW rotation curve generates apparently inconsistent results, e.g., an increase of the baryonic mass leads to increase of the dark matter mass. Furthermore we find that NFW and gNFW profile narrow the total mass range, leading to a possible methodological bias particularly against small MW masses. By using the Einasto profile that is more appropriate to represent cold dark matter haloes, we finally find that the Milky Way slightly decreasing rotation curve favors total mass that can be as small as 2.6 $times 10^{11}$ $M_{odot}$, disregarding any other dynamical tracers further out in the MW. It is inconsistent with values larger than 18 $times 10^{11}$ $M_{odot}$ for any kind of CDM dark-matter halo profiles, under the assumption that stars and gas do not influence the predicted dark matter distribution in the MW. This methodological paper encourages the use of the Einasto profile for characterizing rotation curves with the aim of evaluating their total masses.

Weakly-supervised Graph Meta-learning for Few-shot Node Classification

Graphs are widely used to model the relational structure of data, and the research of graph machine learning (ML) has a wide spectrum of applications ranging from drug design in molecular graphs to friendship recommendation in social networks. Prevailing approaches for graph ML typically require abundant labeled instances in achieving satisfactory results, which is commonly infeasible in real-world scenarios since labeled data for newly emerged concepts (e.g., new categorizations of nodes) on graphs is limited. Though meta-learning has been applied to different few-shot graph learning problems, most existing efforts predominately assume that all the data from those seen classes is gold-labeled, while those methods may lose their efficacy when the seen data is weakly-labeled with severe label noise. As such, we aim to investigate a novel problem of weakly-supervised graph meta-learning for improving the model robustness in terms of knowledge transfer. To achieve this goal, we propose a new graph meta-learning framework -- Graph Hallucination Networks (Meta-GHN) in this paper. Based on a new robustness-enhanced episodic training, Meta-GHN is meta-learned to hallucinate clean node representations from weakly-labeled data and extracts highly transferable meta-knowledge, which enables the model to quickly adapt to unseen tasks with few labeled instances. Extensive experiments demonstrate the superiority of Meta-GHN over existing graph meta-learning studies on the task of weakly-supervised few-shot node classification.

Fairness-aware Personalized Ranking Recommendation via Adversarial Learning

Recommendation algorithms typically build models based on historical user-item interactions (e.g., clicks, likes, or ratings) to provide a personalized ranked list of items. These interactions are often distributed unevenly over different groups of items due to varying user preferences. However, we show that recommendation algorithms can inherit or even amplify this imbalanced distribution, leading to unfair recommendations to item groups. Concretely, we formalize the concepts of ranking-based statistical parity and equal opportunity as two measures of fairness in personalized ranking recommendation for item groups. Then, we empirically show that one of the most widely adopted algorithms -- Bayesian Personalized Ranking -- produces unfair recommendations, which motivates our effort to propose the novel fairness-aware personalized ranking model. The debiased model is able to improve the two proposed fairness metrics while preserving recommendation performance. Experiments on three public datasets show strong fairness improvement of the proposed model versus state-of-the-art alternatives. This is paper is an extended and reorganized version of our SIGIR 2020~cite{zhu2020measuring} paper. In this paper, we re-frame the studied problem as `item recommendation fairness in personalized ranking recommendation systems, and provide more details about the training process of the proposed model and details of experiment setup.

Be More with Less: Hypergraph Attention Networks for Inductive Text Classification

Text classification is a critical research topic with broad applications in natural language processing. Recently, graph neural networks (GNNs) have received increasing attention in the research community and demonstrated their promising results on this canonical task. Despite the success, their performance could be largely jeopardized in practice since they are: (1) unable to capture high-order interaction between words; (2) inefficient to handle large datasets and new documents. To address those issues, in this paper, we propose a principled model -- hypergraph attention networks (HyperGAT), which can obtain more expressive power with less computational consumption for text representation learning. Extensive experiments on various benchmark datasets demonstrate the efficacy of the proposed approach on the text classification task.

A Recent Major Merger Tale for the Closest Giant Elliptical Galaxy Centaurus A

We have used hydrodynamical simulations to model the formation of the closest giant elliptical galaxy, Centaurus A. We find that a single major merger event with a mass ratio up to 1.5, and which has happened ~2 Gyr ago, is able to reproduce many of its properties, including galaxy kinematics, the inner gas disk, stellar halo ages and metallicities, and numerous faint features observed in the halo. The elongated halo shape is mostly made of progenitor residuals deposited by the merger, which also contribute to stellar shells observed in the Centaurus A halo. The current model also reproduces the measured Planetary Nebulae line of sight velocity and their velocity dispersion. Models with small mass ratio and relatively low gas fraction result in a de Vaucouleurs profile distribution, which is consistent with observations and model expectations. A recent merger left imprints in the age distribution that are consistent with the young stellar and Globular Cluster populations (2-4 Gyrs) found within the halo. We conclude that even if not all properties of Centaurus A have been accurately reproduced, a recent major merger has likely occurred to form the Centaurus A galaxy as we observe it at present day.

Graph Prototypical Networks for Few-shot Learning on Attributed Networks

Attributed networks nowadays are ubiquitous in a myriad of high-impact applications, such as social network analysis, financial fraud detection, and drug discovery. As a central analytical task on attributed networks, node classification has received much attention in the research community. In real-world attributed networks, a large portion of node classes only contain limited labeled instances, rendering a long-tail node class distribution. Existing node classification algorithms are unequipped to handle the textit{few-shot} node classes. As a remedy, few-shot learning has attracted a surge of attention in the research community. Yet, few-shot node classification remains a challenging problem as we need to address the following questions: (i) How to extract meta-knowledge from an attributed network for few-shot node classification? (ii) How to identify the informativeness of each labeled instance for building a robust and effective model? To answer these questions, in this paper, we propose a graph meta-learning framework -- Graph Prototypical Networks (GPN). By constructing a pool of semi-supervised node classification tasks to mimic the real test environment, GPN is able to perform textit{meta-learning} on an attributed network and derive a highly generalizable model for handling the target classification task. Extensive experiments demonstrate the superior capability of GPN in few-shot node classification.

A Hierarchical Self-Attentive Model for Recommending User-Generated Item Lists

User-generated item lists are a popular feature of many different platforms. Examples include lists of books on Goodreads, playlists on Spotify and YouTube, collections of images on Pinterest, and lists of answers on question-answer sites like Zhihu. Recommending item lists is critical for increasing user engagement and connecting users to new items, but many approaches are designed for the item-based recommendation, without careful consideration of the complex relationships between items and lists. Hence, in this paper, we propose a novel user-generated list recommendation model called AttList. Two unique features of AttList are careful modeling of (i) hierarchical user preference, which aggregates items to characterize the list that they belong to, and then aggregates these lists to estimate the user preference, naturally fitting into the hierarchical structure of item lists; and (ii) item and list consistency, through a novel self-attentive aggregation layer designed for capturing the consistency of neighboring items and lists to better model user preference. Through experiments over three real-world datasets reflecting different kinds of user-generated item lists, we find that AttList results in significant improvements in NDCG, Precision@k, and Recall@k versus a suite of state-of-the-art baselines. Furthermore, all code and data are available at https://github.com/heyunh2015/AttList.

On the absence of dark matter in dwarf galaxies surrounding the Milky Way

This paper presents an alternative scenario to explain the observed properties of the Milky Way dwarf Spheroidals (MW dSphs). We show that instead of resulting from large amounts of dark matter (DM), the large velocity dispersions observed along their lines of sight can be entirely accounted for by dynamical heating of DM-free systems resulting from MW tidal shocks. Such a regime is expected if the progenitors of the MW dwarfs are infalling gas-dominated galaxies. In this case, gas lost through ram-pressure leads to a strong decrease of self-gravity, a phase during which stars can radially expand, while leaving a gas-free dSph in which tidal shocks can easily develop. The DM content of dSphs is widely derived from the measurement of the dSphs self-gravity acceleration projected along the line of sight. We show that the latter strongly anti-correlates with the dSph distance from the MW, and that it is matched in amplitude by the acceleration caused by MW tidal shocks on DM-free dSphs. If correct, this implies that the MW dSphs would have negligible DM content, putting in question, e.g., their use as targets for DM direct searches, or our understanding of the Local Group mass assembly history. Most of the progenitors of the MW dSphs are likely extremely tiny dIrrs, and deeper observations and more accurate modeling are necessary to infer their properties as well as to derive star formation histories of the faintest dSphs.

Distance and extinction determination for APOGEE stars with Bayesian method

Using a Bayesian technology we derived distances and extinctions for over 100,000 red giant stars observed by the Apache Point Observatory Galactic Evolution Experiment (APOGEE) survey by taking into account spectroscopic constraints from the APOGEE stellar parameters and photometric constraints from 2MASS, as well as a prior knowledge on the Milky Way. Derived distances are compared with those from four other independent methods, the Hipparcos parallaxes, star clusters, APOGEE red clump stars, and asteroseismic distances from APOKASC (Rodrigues et al. 2014) and SAGA Catalogues (Casagrande et al. 2014). These comparisons covers four orders of magnitude in the distance scale from 0.02 kpc to 20 kpc. The results show that our distances agree very well with those from other methods: the mean relative difference between our Bayesian distances and those derived from other methods ranges from -4.2% to +3.6%, and the dispersion ranges from 15% to 25%. The extinctions toward all stars are also derived and compared with those from several other independent methods: the Rayleigh-Jeans Color Excess (RJCE) method, Gonzalezs two-dimensional extinction map, as well as three-dimensional extinction maps and models. The comparisons reveal that, overall, estimated extinctions agree very well, but RJCE tends to overestimate extinctions for cool stars and objects with low logg.