tkan 0.1.2

Temporal KAN model

TKAN: Temporal Kolmogorov-Arnold Networks

TKAN (Temporal Kolmogorov-Arnold Networks) is a neural network architecture designed to enhance multi-horizon time series forecasting. This TensorFlow implementation integrates TKAN as a layer within sequential models, facilitating the use of advanced neural network techniques in practical applications. It is the original implementation of the paper

Installation

Install TKAN directly from PyPI:

pip install tkan

Dependencies are managed using pyproject.toml.

Usage

TKAN can be used within TensorFlow models to handle complex sequential patterns in data. Here is an example that demonstrates how to use TKAN with B-spline activations in a sequential model:

from temporal_kan import TKAN, BSplineActivation
import tensorflow as tf

# Example model using TKAN with B-spline activations
model = tf.keras.Sequential([
    tf.keras.layers.InputLayer(input_shape=X_train.shape[1:]),
    TKAN(activation_funcs=[BSplineActivation(i) for i in range(5)], num_outputs=100, return_sequences=True),
    TKAN(activation_funcs=[BSplineActivation(i) for i in range(1, 4)], num_outputs=100, return_sequences=False),
    tf.keras.layers.Dense(y_train.shape[1], activation='linear')
], name=f'Sequential_Bspline')

Activation Function Flexibility

TKAN layers are highly flexible with regards to activation functions. They can be configured using various types of activations:

• Callable classes: Custom classes like BSplineActivation allow for sophisticated configurations.
• Integers or floats: Specify an initial exponent for a simple power spline activation.
• None: Defaults to BSplineActivation with an order of 3.
• Strings: Utilizes standard TensorFlow activation functions.

Shield:

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.