a b/docs-source/source/model_params.rst 

  1
  
.. currentmodule:: slideflow





  
  

  2
  





  
  

  3
  
.. _model_params:





  
  

  4
  





  
  

  5
  
slideflow.ModelParams





  
  

  6
  
=====================





  
  

  7
  





  
  

  8
  
The :class:`ModelParams` class organizes model and training parameters/hyperparameters and assists with model building.





  
  

  9
  





  
  

  10
  
See :ref:`training` for a detailed look at how to train models.





  
  

  11
  





  
  

  12
  
ModelParams





  
  

  13
  
***********





  
  

  14
  
.. autoclass:: ModelParams





  
  

  15
  
.. autofunction:: slideflow.ModelParams.to_dict





  
  

  16
  
.. autofunction:: slideflow.ModelParams.get_normalizer





  
  

  17
  
.. autofunction:: slideflow.ModelParams.validate





  
  

  18
  
.. autofunction:: slideflow.ModelParams.model_type





  
  

  19
  





  
  

  20
  
Mini-batch balancing





  
  

  21
  
********************





  
  

  22
  





  
  

  23
  
During training, mini-batch balancing can be customized to assist with increasing representation of sparse outcomes or small slides. Five mini-batch balancing methods are available when configuring :class:`slideflow.ModelParams`, set through the parameters ``training_balance`` and ``validation_balance``. These are ``'tile'``, ``'category'``, ``'patient'``, ``'slide'``, and ``'none'``.





  
  

  24
  





  
  

  25
  
If **tile-level balancing** ("tile") is used, tiles will be selected randomly from the population of all extracted tiles.





  
  

  26
  





  
  

  27
  
If **slide-based balancing** ("patient") is used, batches will contain equal representation of images from each slide.





  
  

  28
  





  
  

  29
  
If **patient-based balancing** ("patient") is used, batches will balance image tiles across patients. The balancing is similar to slide-based balancing, except across patients (as each patient may have more than one slide).





  
  

  30
  





  
  

  31
  
If **category-based balancing** ("category") is used, batches will contain equal representation from each outcome category.





  
  

  32
  





  
  

  33
  
If **no balancing** is performed, batches will be assembled by randomly selecting from TFRecords. This is equivalent to slide-based balancing if each slide has its own TFRecord (default behavior).





  
  

  34
  





  
  

  35
  
See :ref:`balancing` for more discussion on sampling and mini-batch balancing.





  
  

  36
  





  
  

  37
  
.. note::





  
  

  38
  





  
  

  39
  
    If you are :ref:`using a Trainer <training_with_trainer>` to train your models, you can further customize the mini-batch balancing strategy by using :meth:`slideflow.Dataset.balance` on your training and/or validation datasets.