perming-1.2.1

fix bugs:

• add Mutipler to inherit parent classes BaseModel, v1.2.0 don't attach it due to work negligence.
• support Linear(1, 1) in model layers to make features' dataset with dimension at (n, 1) possible.
• reduce returns of get length of val_container within __len__ in train_val process when accumulated validation is done.

download:

!pip install perming==1.2.1 # in jupyter
pip install perming==1.2.1 # in cmd

perming-1.2.0

new traits:

• drop install_requires about polars[pandas] which are used to extract, transform, and load dataset in tests.
• replace dependency named sortingx which I developed in more collaborative reverse sorting with bulit-in operator named sorted.
• more comprehensive assertion information prompts and more robust data check in data_loader process, like (n_samples, n_outputs).

download:

!pip install perming==1.2.0 # jupyter
pip install perming==1.2.0 # cmd

perming-1.1.1

fix bugs:

• assert about initial parameters include hidden_layer_sizes, batch_size and learning_rate_init.
• num_epochs and interval > 0 in pre-assertion before set_val_container and train_val process.
• assert str(target.dtype).startswith("float") to ensure target values of regression task.

download:

!pip install perming==1.1.1 # in jupyter
pip install perming==1.1.1 # cmd

same hyper-parameters as v1.0.0

perming-1.0.0

perming: Perceptron Models Are Training on Windows Platform with Default GPU Acceleration.

The supervised learning framework based on perceptron for tabular data.

• always interact with numpy.ndarray and support any well-organized tabular data.
• use perceptron network serving as the basic model to drive machine learning task.
• with object-oriented programming to fully support any supervised learning problems.

download:

!pip install perming==1.0.0 # in jupyter
pip install perming==1.0.0 # cmd

support:

• binary classification
• multiple classification
• regression