Dual-Stream Conscience: From Architecture to Trained Model
In May 2023, Geoffrey Hinton resigned from Google to speak about the risks of AI [1]. In interviews with the New York Times and CBS, he described concerns about AI …
Applying the Brain’s Dual-Library Mechanism to Transformer Architectures
Current large language models (LLMs) concatenate system instructions, historical context, and factual data into the same token sequence. Through successive layers of self-attention, these distinct inputs intertwine — a monolithic …
NELA: Beyond Human Syntax – The Logic of Future Coding Agents
Current software development using AI coding agents rests on a technological paradox. We use coding agents based on transformer architectures with billions of parameters to write code in languages like …
Efficient Kernel Smoother in the ONNX Format
Efficient data processing lies at the heart of modern machine learning. Kernel smoothing, a versatile non-parametric technique, benefits significantly from performance improvements through optimized algorithms and frameworks. In this post, …
Offset and Weights in GLM Regression
Introduction In GLMs, we often encounter scenarios where we need to account for exposure or adjust for certain factors. Both offset and weights play crucial roles in achieving this. Let’s …
Unlocking Stock Market Insights with RiskBERT
Predicting stock prices is akin to the attempts of alchemists in the Middle Ages to transmute lead into gold. Just as alchemists sought to unlock the secrets of transformation, statisticians …
Notes on stochastic count processes with independent non stationary increments
Problem Statement Let thus , we construct a stochastic process as , this kind of random variable are called hypoexponential random variables. We define a counting process such that . …
Generalized Semantic Regression using Contextual Embeddings
In many applications actuaries, data scientists or researchers are confronted with datasets like shown in Table 1. While nominal variables, as in the first or second column, can be used …
Guessing Passwords using an LSTM Network
Introduction A while ago I applied NLP strategies to implement an algorithm that is capable to guess a password. Since then a new method called transformer was developed and successfully …
Expected maxima of a Brownian Motion- Does my stock trading strategy work?
I recently realized that I have a certain habit when it comes to stock trading even though I used to tell myself that “every trading strategy is useless because the …
Estimating the extrema of noisy curves and optimization using spline surface approximation
Estimating maxima and minima of a noisy curve turns out to be very hard and to a large part is still an open question. In this blog post we will discover some strategies …
A WordPress Plugin to embed raw.githubusercontent
Working at my post about the dask cluster, I realized that the code snippets presented in the post will eventually change in my GitHub Repo. I wanted to avoid having …
Building a minimal, cost efficient Dask cluster
In this article we will show a way to do high performance parallel computing at a Kubernetes cluster using task. A primary focus is that we want to archive the …
Frequency-Severity Modeling in consideration of COVID-19 induced effects
This post is supposed to give a brief introduction in Frequency-Severity models. These models are very popular for determine the optimal price for an insurance. We will take a look …
Kernel Regression using the Fast Fourier Transform
1. Setup In a previous post it was shown how to speed up the computation of a kernel density using the Fast Fourier Transform. Conceptually a kernel density is not …
Cluster Monitoring using a ST7789 Display
In the articles Kuberentes at an OrangePi and Setting up the OrangePi it was described how I build my toy cluster. Meanwhile the cluster received some updates. Two more nodes …
Fast Kernel Density Estimation using the Fast Fourier Transform
1. Setup This Post is about how to speed up the computation kernel density estimators using the FFT (Fast Fourier Transform). Let be be a random sample drawn from an …
Non-Linear Classification Methods in Spark
In a previous post I covered how to apply classical linear estimators like support vector machines or logistic regression to a non-linear dataset using the kernel method. This article can …
Kubernetes at an OrangePi
1. Install k3s at the OrangePI In a previous article I explained how to get spark running at an OrangePi to create a toy computing-cluster. If you look at this …
An AI with less than 200 lines of code
In the last two articles we covered the topics “How to teach a computer gamerules” and “The Multiarmed Bandit Problem”. Indeed these two articles where intended to be an introduction …