Category: Analytics
-
Personal metalearning
Having 30+ years history in a knowledge domain, it’s a surprise if sudden level ups start appearing. Metalearning seems to have done that for my chess, after a longish warm up period of getting tuned in for the game. LLMs are an inspiring success story, and AI is sometimes giving feelings of “getting everything solved”.…
-
GPT Philosopher II
There’s been some positive changes and challenges for me on my other projects since the last autumn, explaining the quietness here. Now things have progressed so that I was able to find time on my home programming desk. Large language model APIs are on the focus again. Earlier this year I was running some tests…
-
Solving levels
The fun puzzle game Active Neurons 3 got me to Nethack feelings when solving the game levels using Python. Any point in the game, you can move the player figure towards any of the four main directions, and the movement stops only after bumbing into an element like a wall. There’s a beginning location, and…
-
Database and tools
I had been thinking about having a dedicated relational database server at home, and found an unused old PC that was laying around and fit for the purpose. Changing to Ubuntu and installing the chosen open source database was quite fast. Then there was setting up a secure connection between the server and my main…
-
Underground data
Public transit offers some nice open data sets. I was searching for subway data, and found out that New York releases a large data batch of metro arrival times every morning. Not knowing much on the NYC metro, there was some research to be done, going through the metro map, route time tables, station locations…
-
Logs on logic
Processes write logs, and sometimes you have the data without access to the process. I was trying out generating log data using number of repeating loggers defined by graphs. The next step I ran some statistics on the log, and depending on the parameters of the log data generator, it was more or less easy…
-
GPT Philosopher
Currently my favourite desktop app is Obsidian. It’s a software for personal knowledge management. Giving the light experience of a plain text editor, it has features like super easy hypertext functionality, knowlege graph, and PowerPoint type canvas. Few days ago I decided to try automating philosophical knowledge transfer from ChatGPT to Obsidian using Python. The…
-
Outliers?
Happy holidays! I started working on an anomaly detection system that thought I need. The first step was to have data downloads from an API, and some functions storing data for later use. The API returns data in JSON format, but here you effectively deal with a list of tuples. Instead of using plain tuples,…
-
Yeah it’s that thing
The idea of prompt engineering is to verbalize the required application structure for the AI. Sometimes it feels that a faster way to solve problems is only reacting to the situation and producing a solution. Choosing to verbalize is often an attempt to make the structure explicit, maybe using some set of methods like computational…
-
Data apps
I’ve been working on data apps recently. The idea is to pick a data source X, and create an app that captures the basic analytic functions that you want to run on it. What it means in practice is having a data model, data imports and exports, some automated analytics, reports, and a graphical user…