tallamjr/astronet

Issues

• Minor modifications in relation to viva corrections

  #99 opened 2 years ago by tallamjr
  0

• Update functionality to allow for processing of ELAsTiCC data

  #98 opened 2 years ago by tallamjr
  0

• Lazy loading of large numpy data files

  #96 opened 3 years ago by tallamjr
  0

• [META] Road to `v1.0.0`

  #87 opened 3 years ago by tallamjr
  0

• Fix pytest-xdist with session-scoped fixture

  #94 opened 3 years ago by tallamjr
  0

• MINOR: Refactor code layout now that a "new" kind of model exists: `tinho`

  #92 opened 3 years ago by tallamjr
  0

• Fix multi-gpu training

  #90 opened 3 years ago by tallamjr
  1

• Apply model compression techniques for improved throughput

  #89 opened 3 years ago by tallamjr
  0

• Re-implement GPs

  #77 opened 3 years ago by tallamjr
  2

• Compare runtimes of `atx` vs `t2`

  #79 opened 3 years ago by tallamjr
  4

• Update docstrings

  #24 opened 4 years ago by tallamjr
  1

• [RFC] Improve data reformatting pipeline

  #86 opened 3 years ago by tallamjr
  0

• Experiment with `atx` for FINK

  #84 opened 3 years ago by tallamjr
  1

• Clean codebase

  #69 opened 3 years ago by tallamjr
  2

• Revisit implementation of Multi-head attention

  #38 opened 4 years ago by tallamjr
  1

• Broker integration

  #81 opened 3 years ago by tallamjr
  1

• Run `atx` on Myriad

  #78 opened 3 years ago by tallamjr
  0

• Optimize for accuracy for MTS comparison

  #76 opened 4 years ago by tallamjr
  2

• Clean CAM code

  #68 opened 4 years ago by tallamjr
  2

• Conduct ablation studies

  #67 opened 4 years ago by tallamjr
  2

• Run without redshift

  #72 opened 4 years ago by tallamjr
  1

• Re-do concatenation of features to L or M before Conv layer

  #70 opened 4 years ago by tallamjr
  1

• Create dataset for full set of transients in PLAsTiCC dataset

  #65 opened 4 years ago by tallamjr
  1

• [MAJOR] Revisit model architecture

  #35 opened 4 years ago by tallamjr
  0

• Expand hyperparameter space of `t2` model

  #62 opened 4 years ago by tallamjr
  0

• Improve plots for explainability of `t2` model and the attention mechanism

  #63 opened 4 years ago by tallamjr
  0

• Balance classes with GP augmentation

  #32 opened 4 years ago by tallamjr
  0

• Include `pool_size` as hyperparameter to optimise over

  #60 opened 4 years ago by tallamjr
  1

• Add redshift information to model

  #31 opened 4 years ago by tallamjr
  2

• Add Codecov report via Github actions

  #48 opened 4 years ago by tallamjr
  1

• [MAJOR] Implement snX: Supernova Xception

  #58 opened 4 years ago by tallamjr
  3

• [BENCHMARK] Compare results to `snmachine`

  #44 opened 4 years ago by tallamjr
  3

• Evaluation notebook

  #39 opened 4 years ago by tallamjr
  1

• Visualise Self-Attention weights response

  #45 opened 4 years ago by tallamjr
  1

• [BENCHMARK] Compare results to `dl-4-tsc` on the MTS dataset

  #52 opened 4 years ago by tallamjr
  1

• Custom loss function error: TypeError: __init__() takes from 1 to 2 positional arguments but 3 were given

  #54 opened 4 years ago by tallamjr
  2

• AttributeError: module 'tensorflow._api.v2.experimental' has no attribute 'numpy'

  #56 opened 4 years ago by tallamjr
  1

• Add functionality to compare results against Fawaz et al.

  #51 opened 4 years ago by tallamjr
  0

• Refactor how to load datasets

  #53 opened 4 years ago by tallamjr
  1

• Determine optimal mini batch size

  #55 opened 4 years ago by tallamjr
  2

• Add callbacks to monitor training progress and prevent overfitting

  #47 opened 4 years ago by tallamjr
  1

• Refactor `def train_val_test_split(df, cols):` to be more configurable

  #43 opened 4 years ago by tallamjr
  0

• Stack Encoding Layers to be in line with Vaswani paper

  #37 opened 4 years ago by tallamjr
  2

• Reimplement custom log loss as a tf.keras object

  #41 opened 4 years ago by tallamjr
  0

• [MR/35] Implement `PositionalEncoding` class

  #36 opened 4 years ago by tallamjr
  3

• Investigate large window size, timesteps=100; steps=100

  #33 opened 4 years ago by tallamjr
  1

• Re-train PLAsTiCC model with 'weighted' Log-Loss metric and loss for optimisation

  #30 opened 4 years ago by tallamjr
  2

• Update results and analysis with hyperparamters tuned on validation set instead of test

  #26 opened 4 years ago by tallamjr
  0

• Reference to other files within code is all relative to the __file__ being run

  #27 opened 4 years ago by tallamjr
  0

• [feature/20/plasticc] Investigate options for Gaussian Process regression for PLAsTiCC data

  #21 opened 4 years ago by tallamjr
  2