vectorflow

Vectorflow is a lightweight neural network library for sparse data.

Modules

dataset
module vectorflow.dataset

This module provides utility classes to iterate over data.

layers
module vectorflow.layers

Implementation of the various computational layers.

losses
module vectorflow.losses

The library supports already implemented loss functions, as well as a callback-based way to specify a custom loss.

math
module vectorflow.math

Low-level math primitives.

ml
module vectorflow.ml

Utility functions for ML-related tasks.

monitor
module vectorflow.monitor

Internal abstraction to report progress during training.

neurallayer
module vectorflow.neurallayer

Base class for all layers.

neuralnet
module vectorflow.neuralnet

NeuralNet is the main abstraction of vectorflow.

optimizers
module vectorflow.optimizers

Implementation of different stochastic optimizers.

regularizers
module vectorflow.regularizers

Implementation of standard regularizers for Linear layer.

serde
module vectorflow.serde

Internal serialization functions for automatic neural net ser/deser.

utils
module vectorflow.utils

Utility functions used across the library.

Meta

Authors

Benoit Rostykus