File size: 2,930 Bytes
6b00fe9 8f782b4 6b00fe9 8f782b4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
---
license: mit
tags:
- time series
- astrophysics
- pretraining
- connect-later
size_categories:
- 100K<n<1M
---
# AstroClassification and Redshifts Datasets
<!-- Provide a quick summary of the dataset. -->
This dataset was used for the AstroClassification and Redshifts introduced in [Connect Later: Improving Fine-tuning for Robustness with Targeted Augmentations](). This is a dataset of simulated astronomical time-series (e.g., supernovae, active galactic nuclei), and the task is to classify the object type (AstroClassification) or predict the object's redshift (Redshifts).
- **Repository:** https://github.com/helenqu/connect-later
- **Paper:** will be updated
- **Point of Contact: Helen Qu (<[email protected]>)**
## Dataset Structure
<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
- **object_id**: unique object identifier
- **times_wv**: 2D array of shape (N, 2) containing the observation times (modified Julian days, MJD) and filter (wavelength in nm) for each observation, N=number of observations
- **lightcurve**: 2D array of shape (N, 2) containing the flux (arbitrary units) and flux error for each observation
- **label**: integer representing the class of the object (see below for details)
- **redshift**: redshift of the object
## Dataset Creation
### Source Data
This is a modified version of the dataset from the 2018 Photometric LSST Astronomical Time-Series Classification Challenge (PLAsTiCC) Kaggle competition
The original Kaggle competition can be found [here](https://www.kaggle.com/c/PLAsTiCC-2018). [This note](https://arxiv.org/abs/1810.00001) from the competition describes the dataset in detail. Astronomers may be interested in [this paper](https://arxiv.org/abs/1903.11756) describing the simulations used to generate the data.
- **Train**: 80% of the original PLAsTiCC training set augmented using the redshifting targeted augmentation described in the Connect Later paper
- **Validation**: Remaining 20% of the original PLAsTiCC training set, *not* augmented or modified
- **Test**: Subset of 10,000 objects randomly selected from the PLAsTiCC test set
### Object Types
```
0: microlens-single
1: tidal disruption event (TDE)
2: eclipsing binary (EB)
3: type II supernova (SNII)
4: peculiar type Ia supernova (SNIax)
5: Mira variable
6: type Ibc supernova(SNIbc)
7: kilonova (KN)
8: M-dwarf
9: peculiar type Ia supernova (SNIa-91bg)
10: active galactic nuclei (AGN)
11: type Ia supernova (SNIa)
12: RR-Lyrae (RRL)
13: superluminous supernova (SLSN-I)
14: 5 "anomalous" types that are not present in training set: microlens-binary, intermediate luminosity optical transient (ILOT), calcium-rich transient (CaRT), pair instability supernova (PISN), microlens-string
```
## Citation
will be updated |