KANLab: TensorFlow Playground for Kolmogorov-Arnold Networks

Lukas Niehaus; Robin Rawiel; Ulf Krumnack

cs.LGOct 14, 2026

KANLab: TensorFlow Playground for Kolmogorov-Arnold Networks

Authors: Lukas Niehaus, Robin Rawiel, Ulf Krumnack

Organizations: University Osnabrueck · University of Applied Sciences Osnabrueck

Abstract

KANLab is a browser-based visualization tool designed to interactively teach the core concepts of Kolmogorov-Arnold Networks (KAN), forked from the widely successful TensorFlow Playground. To maximize educational clarity, KANLab omits residual connections and learnable scaling parameters found in modern KAN implementations, ensuring that displayed B-spline activation curves directly correspond to their underlying trainable control points. Users can inspect and modify activation functions, initialization schemes and observe gradient descent correcting manual disruptions. Furthermore, user can explore phenomena such as activation variance, interactive pruning, and re-/training within a compact 2D machine learning setting. KANLab is open-source and available at https://kanlab.app/

Figures & tables

Figure 1: Main view of KANLab, with a two layer kan that has 2 input features ( $P$ ), 2 hidden nodes ( $Q$ ) and 1 output Node, represented by squares. In between these we visualize the green B-spline activation functions with rectangles. Left and right of the rectangles gray dashed lines represent the variance of the incoming (left) and outgoing (right) activation variance by contrasting levels of thickness.

Figure 2: Visualization of an activation function parameterized by a B-spline and its effect on activations. The green curve shows the current B-spline activation, while the orange points denote its learnable control points. The top histogram depicts the distribution of incoming activations, and the right histogram shows the distribution of outgoing activations after applying the activation function. Adjusting the orange control points changes the shape of the B-spline and, consequently, the mapping from inputs to outputs.

Kolmogorov-Arnold Networks

Kolmogorov-Arnold Networks (KANs) learn compositional functions through trainable one-dimensional edge transformations for approximation and interpretable modeling.

See all 125 papers

InKAN: B-Spline KANs via Truncated Power Form
Sep 2, 2026Naveen MysoreEfficient Neural Network Inference Neural Network Activation Functions
KANs need curvature: penalties for compositional smoothness
May 4, 2026James BagrowNeural Network Activation Functions Neural Network Interpretability
Mathematics & Statistics, University of Vermont, Burlington, VT, United States · Vermont Complex Systems Center, University of Vermont, Burlington, VT, United States
Geometry-Constrained Kolmogorov-Arnold Networks: Learning Edge Geometry via Banach Duality
Aug 26, 2026K S Sesh KumarKolmogorov-Arnold Networks Symbolic Regression
Brevan Howard Centre for Financial Analysis Imperial Business School

Sep 2, 2026J/K move · Enter open · S save

InKAN: B-Spline KANs via Truncated Power Form

Naveen Mysore

Efficient Neural Network Inference Neural Network Activation Functions Kolmogorov-Arnold Networks

Open paper AskPDF arXiv

Kolmogorov-Arnold Networks (KANs) place learnable B-spline activations on network edges rather than fixed activations on nodes. The standard Cox-de Boor recursion evaluates these activations through

k

sequential passes for degree-

k

splines, consuming over 90% of forward-pass time. InKAN replaces this recursion with the truncated power form, a classical result from approximation theory that expresses each uniform cubic B-spline as five

(x)_+^3

terms at shifted knot positions. The resulting expression computes exact B-spline basis values: the same mathematical function as the Cox-de Boor recursion, evaluated without sequential passes. This paper documents three contributions: (1) an implementation structured for torch . compile fusion, eliminating recursion, span lookup, and scatter-gather operations; (2) a bounded-coordinate evaluation that clamps the normalized input to

[0, k{+}1]

, preventing the growth of cancellation error at large off-support coordinates; and (3) an open-source package (pip install inkan). In the tested configurations, InKAN has 2.8--3.5

\times

lower forward-pass latency than the Cox-de Boor recursion. Partition-of-unity errors remain below

10^{-5}

for grid sizes up to 200.

Abstract

Kolmogorov-Arnold Networks

Explore similar work

InKAN: B-Spline KANs via Truncated Power Form

KANs need curvature: penalties for compositional smoothness

Geometry-Constrained Kolmogorov-Arnold Networks: Learning Edge Geometry via Banach Duality