Abstract:Astrophysical transient phenomena are traditionally classified spectroscopically in a hierarchical taxonomy; however, this graph structure is currently not utilized in neural net-based photometric classifiers for time-domain astrophysics. Instead, independent classifiers are trained for different tiers of classified data, and events are excluded if they fall outside of these well-defined but flat classification schemes. Here, we introduce a weighted hierarchical cross-entropy objective function for classification of astrophysical transients. Our method allows users to directly build and use physics- or observationally-motivated tree-based taxonomies. Our weighted hierarchical cross-entropy loss directly uses this graph to accurately classify all targets into any node of the tree, re-weighting imbalanced classes. We test our novel loss on a set of variable stars and extragalactic transients from the Zwicky Transient Facility, showing that we can achieve similar performance to fine-tuned classifiers with the advantage of notably more flexibility in downstream classification tasks.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that the current neural network methods used for photometric classification in time - domain astronomy fail to fully utilize the hierarchical classification structure of astrophysical transients. Specifically, traditional spectral classification methods adopt a flattened classification scheme when dealing with astrophysical transients, which results in a large amount of data being excluded because these data cannot be well - adapted to the predefined classification system. In addition, existing classification methods usually need to train independent classifiers for data at different levels, which not only increases the computational cost but also may lead to less flexible and accurate classification results. To address these problems, the author proposes a new Weighted Hierarchical Cross - Entropy loss function (WHXE) to directly utilize the tree - like classification structure based on physical or observational motives. WHXE improves the flexibility and accuracy of classification by re - weighting unbalanced classes and allowing all targets to be classified into any node in the tree. ### Main problem summary: 1. **Lack of utilization of hierarchical structure**: Existing methods do not fully utilize the natural hierarchical classification structure of astrophysical transients. 2. **Low data utilization rate**: Due to the limitations of the classification system, a large amount of data is excluded. 3. **Complexity of multi - level classifiers**: It is necessary to train multiple independent classifiers for data at different levels, which increases the computational cost. ### Solution: - **Weighted Hierarchical Cross - Entropy loss function (WHXE)**: By introducing hierarchical structure and class weights, the classifier can handle unbalanced data sets more flexibly and fully utilize all available data. ### Formula representation: The weighted hierarchical cross - entropy loss function is defined as follows: \[ L_{\text{WHXE}}(p, c)=-\sum_{h = 0}^{H - 1}W(c(h))\lambda(c(h))\log p(c(h)|c(h + 1)) \] where: - \(W(c(h))\) is the class weight, which is used to re - weight the class proportion of each object. - \(\lambda(c(h))\) is the second weight, which is used to emphasize different levels of the tree and is defined as: \[ \lambda(c)=\exp(-\alpha h(c)) \] where \(\alpha\) is a hyperparameter, and a larger value of \(\alpha\) will place more emphasis on the top - level classification of the hierarchical structure. Through this method, the author can significantly improve data utilization and classification flexibility while maintaining classification performance.

Hierarchical Cross-entropy Loss for Classification of Astrophysical Transients

Classifying Image Sequences of Astronomical Transients with Deep Neural Networks

Nonparametric Transient Classification using Adaptive Wavelets

Transient-optimised real-bogus classification with Bayesian Convolutional Neural Networks -- sifting the GOTO candidate stream

Optical Transient Object Classification in Wide Field Small Aperture Telescopes with Neural Networks

A Classifier-Based Approach to Multi-Class Anomaly Detection for Astronomical Transients

Light curve classification with recurrent neural networks for GOTO: dealing with imbalanced data

First Impressions: Early-Time Classification of Supernovae using Host Galaxy Information and Shallow Learning

TLW: A Real-Time Light Curve Classification Algorithm for Transients Based on Machine Learning

Identifying Light-curve Signals with a Deep-learning-based Object Detection Algorithm. II. A General Light-curve Classification Framework

Photometric classification of HSC transients using machine learning

Real-Time Detection of Anomalies in Large-Scale Transient Surveys

Paying Attention to Astronomical Transients: Introducing the Time-series Transformer for Photometric Classification

Scalable End-to-end Recurrent Neural Network for Variable star classification

Identifying Light-curve Signals with a Deep Learning Based Object Detection Algorithm. II. A General Light Curve Classification Framework

Transient Classification in low SNR Gravitational Wave data using Deep Learning

Unsupervised Transient Light Curve Analysis Via Hierarchical Bayesian Inference

Real-time Detection of Anomalies in Multivariate Time Series of Astronomical Data

Avocado: Photometric Classification of Astronomical Transients with Gaussian Process Augmentation

RAPID: Early Classification of Explosive Transients using Deep Learning

What's the Difference? The potential for Convolutional Neural Networks for transient detection without template subtraction