deep learning - Marcus D. R. Klarqvist

Optimizers in deep learning

Gradient descent and its many variations are the de facto algorithms to optimize modern deep learning models. Deep learning libraries offers a multitude of these optimization algorithms and often remain enigmatic to most users and are treated as black boxes due to a lack of accessible explanations.

machine learning

Interesting Functions for Testing Optimization Methods

Optimization is quintessential to machine learning and deep learning, with objective functions serving to represent the problem's objectives and constraints. These functions vary from straightforward quadratics to intricate landscapes that mirror real-world challenges.

machine learning

Regularization in Machine Learning: Part 1

Dive deep into model generalizability, bias-variance trade-offs, and the art of regularization. Learn about L2 and L1 penalties and automatic feature selection. Apply these techniques to a real-world use-case!

deep learning

AI in healthcare: electrocardiograms - Part 1

Dive into the world of healthcare AI with this comprehensive guide! No math, just pure practical insights. Skip the math and delve straight into practical engineering. Explore electrocardiogram data, from data prep to visualization. Discover why data analysis is the backbone of machine learning.

deep learning

Automatic differentiation in deep learning: Part 2

Dive into the world of autodiff as we explore the relationship between dual numbers and derivatives. Discover how to leverage this connection to compute gradients effortlessly. Then build a Python engine to explore forward- and reverse-mode autodiff and dissecting function compositions.

deep learning

Automatic differentiation in deep learning

Dive into the fascinating world of automatic differentiation! Say goodbye to tedious hand calculations. Learn about dual numbers, automatic differentiation methods, and implement your own Python program from scratch. Discover the power and efficiency of modern deep learning techniques firsthand.