Search

Temporal set prediction is becoming increasingly important as many companies employ recommender systems in their online businesses, e.g., personalized purchase prediction of shopping baskets. While most previous techniques have focused on leveraging a user's history, the study of combining it with others' histories remains untapped potential. This paper proposes Global-Local Item Embedding (GLOIE) that learns to utilize the temporal properties of sets across whole users as well as within a user by coining the names as global and local information to distinguish the two temporal patterns. GLOIE uses Variational Autoencoder (VAE) and dynamic graph-based model to capture global and local information and then applies attention to integrate resulting item embeddings. Additionally, we propose to use Tweedie output for the decoder of VAE as it can easily model zero-inflated and long-tailed distribution, which is more suitable for several real-world data distributions than Gaussian or multinomial counterparts. When evaluated on three public benchmarks, our algorithm consistently outperforms previous state-of-the-art methods in most ranking metrics., Comment: 8 pages, 3 figures. To appear in RecSys 2021 LBR

We present ${\tt nimbus}$ : a hierarchical Bayesian framework to infer the intrinsic luminosity parameters of kilonovae (KNe) associated with gravitational-wave (GW) events, based purely on non-detections. This framework makes use of GW 3-D distance information and electromagnetic upper limits from multiple surveys for multiple events, and self-consistently accounts for finite sky-coverage and probability of astrophysical origin. The framework is agnostic to the brightness evolution assumed and can account for multiple electromagnetic passbands simultaneously. Our analyses highlight the importance of accounting for model selection effects, especially in the context of non-detections. We show our methodology using a simple, two-parameter linear brightness model, taking the follow-up of GW190425 with the Zwicky Transient Facility (ZTF) as a single-event test case for two different prior choices of model parameters -- (i) uniform/uninformative priors and (ii) astrophysical priors based on surrogate models of Monte Carlo radiative transfer simulations of KNe. We present results under the assumption that the KN is within the searched region to demonstrate functionality and the importance of prior choice. Our results show consistency with ${\tt simsurvey}$ -- an astronomical survey simulation tool used previously in the literature to constrain the population of KNe. While our results based on uniform priors strongly constrain the parameter space, those based on astrophysical priors are largely uninformative, highlighting the need for deeper constraints. Future studies with multiple events having electromagnetic follow-up from multiple surveys should make it possible to constrain the KN population further., Comment: 18 pages, 5 figures, accepted for publication in ApJ

General-purpose representation learning through large-scale pre-training has shown promising results in the various machine learning fields. For an e-commerce domain, the objective of general-purpose, i.e., one for all, representations would be efficient applications for extensive downstream tasks such as user profiling, targeting, and recommendation tasks. In this paper, we systematically compare the generalizability of two learning strategies, i.e., transfer learning through the proposed model, ShopperBERT, vs. learning from scratch. ShopperBERT learns nine pretext tasks with 79.2M parameters from 0.8B user behaviors collected over two years to produce user embeddings. As a result, the MLPs that employ our embedding method outperform more complex models trained from scratch for five out of six tasks. Specifically, the pre-trained embeddings have superiority over the task-specific supervised features and the strong baselines, which learn the auxiliary dataset for the cold-start problem. We also show the computational efficiency and embedding visualization of the pre-trained features.

In a search for eclipsing white dwarfs using the Zwicky Transient Facility lightcurves, we identified a deep eclipsing white dwarf with a dark, substellar companion. The lack of an infrared excess and an orbital period of 10 hours made this a potential exoplanet candidate. We obtained high-speed photometry and radial velocity measurements to characterize the system. The white dwarf has a mass of $0.50\pm0.02\,\mathrm{M_{\odot}}$ and a temperature of $10900\pm200\,$K. The companion has a mass of $0.059\pm0.004\,\mathrm{M_{\odot}}$ and a small radius of $0.0783\pm0.0013\,\mathrm{R_{\odot}}$. It is one of the smallest transiting brown dwarfs known and likely old, $\gtrsim 8\,$Gyr. The ZTF discovery efficiency of substellar objects transiting white dwarfs is limited by the number of epochs and as ZTF continues to collect data we expect to find more of these systems. This will allow us to measure period and mass distributions and allows us to understand the formation channels of white dwarfs with substellar companions., Comment: submitted, comments welcome

We present SNIascore, a deep-learning based method for spectroscopic classification of thermonuclear supernovae (SNe Ia) based on very low-resolution (R $\sim100$) data. The goal of SNIascore is fully automated classification of SNe Ia with a very low false-positive rate (FPR) so that human intervention can be greatly reduced in large-scale SN classification efforts, such as that undertaken by the public Zwicky Transient Facility (ZTF) Bright Transient Survey (BTS). We utilize a recurrent neural network (RNN) architecture with a combination of bidirectional long short-term memory and gated recurrent unit layers. SNIascore achieves a $<0.6\%$ FPR while classifying up to $90\%$ of the low-resolution SN Ia spectra obtained by the BTS. SNIascore simultaneously performs binary classification and predicts the redshifts of secure SNe Ia via regression (with a typical uncertainty of $<0.005$ in the range from $z = 0.01$ to $z = 0.12$). For the magnitude-limited ZTF BTS survey ($\approx70\%$ SNe Ia), deploying SNIascore reduces the amount of spectra in need of human classification or confirmation by $\approx60\%$. Furthermore, SNIascore allows SN Ia classifications to be automatically announced in real-time to the public immediately following a finished observation during the night., Comment: 12 pages, 5 figures, 2 tables, accepted for publication in ApJL

In recent years, graph neural networks (GNNs) have been widely adopted in the representation learning of graph-structured data and provided state-of-the-art performance in various applications such as link prediction, node classification, and recommendation. Motivated by recent advances of self-supervision for representation learning in natural language processing and computer vision, self-supervised learning has been recently studied to leverage unlabeled graph-structured data. However, employing self-supervision tasks as auxiliary tasks to assist a primary task has been less explored in the literature on graphs. In this paper, we propose a novel self-supervised auxiliary learning framework to effectively learn graph neural networks. Moreover, this work is the first study showing that a meta-path prediction is beneficial as a self-supervised auxiliary task for heterogeneous graphs. Our method is learning to learn a primary task with various auxiliary tasks to improve generalization performance. The proposed method identifies an effective combination of auxiliary tasks and automatically balances them to improve the primary task. Our methods can be applied to any graph neural network in a plug-in manner without manual labeling or additional data. Also, it can be extended to any other auxiliary tasks. Our experiments demonstrate that the proposed method consistently improves the performance of node classification and link prediction., Comment: Submitted to TPAMI, 14pages. arXiv admin note: substantial text overlap with arXiv:2007.08294

Assessing advertisements, specifically on the basis of user preferences and ad quality, is crucial to the marketing industry. Although recent studies have attempted to use deep neural networks for this purpose, these studies have not utilized image-related auxiliary attributes, which include embedded text frequently found in ad images. We, therefore, investigated the influence of these attributes on ad image preferences. First, we analyzed large-scale real-world ad log data and, based on our findings, proposed a novel multi-step modality fusion network (M2FN) that determines advertising images likely to appeal to user preferences. Our method utilizes auxiliary attributes through multiple steps in the network, which include conditional batch normalization-based low-level fusion and attention-based high-level fusion. We verified M2FN on the AVA dataset, which is widely used for aesthetic image assessment, and then demonstrated that M2FN can achieve state-of-the-art performance in preference prediction using a real-world ad dataset with rich auxiliary attributes., Comment: published in Applied Soft Computing