Notes from the Deep End

Notes from the Deep End

Sign in Subscribe

Feature image: Translated digits

Building Controllable Latent Representations with Typed Autoencoders

Earlier we discussed latent spaces (in the context of AE and VAE). Let's explore a bit more deeply how to actually make the latent space useful by explicitly defining semantics of named coordinates. So far latent spaces have just been vector spaces with no clear inherent semantic structure.

From Autoencoders to VAE: Smooth Latent Spaces

From Autoencoders to VAE: Smooth Latent Spaces

In the previous article we noticed that with autoencoders the learned latent space was very sparse. The locations (points) in latent space that can produce any meaningful images are few and far between. When building generative models this is a problem. We wouldn't want a model that produces

Autoencoders: The Foundation of Generative AI

Autoencoders: The Foundation of Generative AI

The world of deep learning has evolved dramatically over the last decade or two. While on my post-graduate studies around 2010, the neural networks were mostly a mathematical model that was way too expensive to train for many practical tasks. Todays hardware evolution and the innovations in network architectures have

The Potential and Threat of Quantum Computing in Warfare

The Potential and Threat of Quantum Computing in Warfare

[This text was originally published in Viestimies-journal. This version is slightly updated and translated to english.] This article briefly reviews the nature of quantum computing and explores the opportunities and threats it brings. The perspective focuses on information security and the vision of warfare evolution. Theoretical Background The foundation of

NEW TECHNOLOGY IN THE BATTLEFIELD

NEW TECHNOLOGY IN THE BATTLEFIELD

This article was originally published in Viestimies-journal in Finnish. This is translation to English with some minor changes. The pictures are from the RAND publication Distributed Kill Chains. Leveraging New Technology In The Battlefield Technological development is visible on the todays battlefield. Increased sensors enable new ways of gathering information;

Intent-Aware Semantic Search: An Architectural Approach

Intent-Aware Semantic Search: An Architectural Approach

The problem setup I worked in a startup focusing on open-source data-platforms. As you might guess, building visibility is one of the first things you need to do in startups, and this was not different. As the topic we were working on was hot, I thought it would be rather

Dimension reductions: t-SNE

Dimension reductions: t-SNE

T-SNEis well-known dimensionality reduction algorithm particularly suited for the visualization of high-dimensional datasets. This article goes through the mathematics behind t-SNE and provides a simple, pedagogical Python-implementation on the way. T-SNE stands for t-Distributed Stochastic Neighbor Embedding and it is a non-linear method. 1. It is particularly good at preserving

How Private Tech Surpassed Defense's Technology Advantage

How Private Tech Surpassed Defense's Technology Advantage

The conventional wisdom held that only defense could marshal the resources needed for truly groundbreaking innovation, but this has changed.

CRYSTALS - Kyber Compression and KEM

CRYSTALS - Kyber Compression and KEM

Post-quantum algorithm CRYSTALS Kyber KEM and variable compression explained.

CRYSTALS - Kyber core

CRYSTALS - Kyber core

We introduce the CRYSTALS Kyber algorithms core implementation with math and Python implementation. The presentation relies on the previous article laying the foundations.

CRYSTALS - The Gentle Introduction

CRYSTALS - The Gentle Introduction

An introduction to mathematics used in post-quantum encryption algorithm(s) CRYSTALS. We use Python to implement the machinery as we go.

Technology as a Strategy- Lessons from the Defense Industry

Technology as a Strategy- Lessons from the Defense Industry

For decades, the defense industry has let tech carve the path and trusted results to follow. It’s a perspective businesses could learn from.