Loading animation

Loading...

Kiuyha

Articles List

(4)

Can KANs Replace MLPs? Testing B-Spline, Gaussian, and Fourier Variants

For years, deep learning has mainly focused on the Multi-Layer Perceptron architecture as the primary approach. This recipe is already default in our mind, and many people think that this is the only way a neural network works. But what if we flipped that logic entirely? Enter the Kolmogorov-Arnold...

kiuyha
Jan 13, 2026, 12:15 PM
ai-benchmarking
artificial-intelligence
kolmogorov-arnold-network
deep-learning
pytorch
You Cry about Overfitting. I call it the Warm-Up

If you're already in the machine learning field long enough, you should know about the Golden Rule of training a model. It is a Bias-Variance Trade-off. We always treat it like divine revelation. For a long time, I never really cared about the theory since I was too busy building things, and I only...

kiuyha
Dec 28, 2025, 3:01 AM
deep-learning
modeling
double-descent
machine-learning
Decision Tree from Scratch

In this article, we’ll walk through a decision tree. A supervised machine learning that is labeled as the “white box,” which means it is easy to understand what happens behind the scenes. Not only will I explain how it works in simple terms, but I will also make it from scratch without any external...

kiuyha
Nov 2, 2025, 11:05 AM
artificial-intelligence
machine-learning
decision-tree
sklearn
ai-implementation
Forget sklearn for a Moment: Understanding KNN by Coding It Yourself

One of the simplest machine learning algorithms is KNN, which stands for K-Nearest Neighbors. It’s a supervised machine learning, meaning the dataset is labeled. If you are new to machine learning, you may wonder how straightforward it is. I can say for ML that being able to do both regression and ...

kiuyha
Oct 17, 2025, 8:05 AM
knn-algorithm
machine-learning
artificial-intelligence
supervised-learning
classification