The Analytic Garden

I didn't think this would work. I was right. TimeGPT-1  from Nixtla is a pretrained transformer model for time series forecasting and anomaly detection. It was trained on thousands of timeseries. The idea is that just like text, time series have recognizable patterns and an GPT can leverage these patterns to predict the next number in the series just as an LLM predicts the next token in a series of tokens. I was interested in seeing how it performs on recent temperature anomaly data. Temperature anomaly is the difference of a temperature from a reference value. Here's an example. The monthly temperature anomaly data for the plot was downloaded from  https://www.metoffice.gov.uk/hadobs/hadcrut5/data/HadCRUT.5.0.2.0/download.html .  The plot shows the temperature anomaly for June of each year. The data represents temperature anomalies (deg C) relative to the average global temperature from 1961-1990.  Notice the jump in the temperature anomaly at the last two poi...

  Generated by Janus Pro 7B  https://www.yeschat.ai/features/janus-pro I have trouble with PyMC . PyMC is a probabilistic programming library for Python for building Bayesian models. Whenever I try to use it, I end up searching the web or asking an LLM to help with current syntax. It seems that there are two active versions, PyMC3 and PyMC5. The apis for the versions differ. PyMC5 is still under development. My trouble is that LLMs confuse the two and mix up the syntax of the versions. Searching the web leads to more confusion because many posts don't indicate which version they are referring to.   ChatGPT o3-mini has just been released. I wanted to see how DeepSeek-R1 stacked against it for a simple programming task using PyMC. I gave both the same simple prompt. generate python code to locate change points in a time series using pymc5 ChatGPT o3-mini did a reasonable job. The code worked without modifications. It generated some nice graphs using ArviZ to illu...

In a previous post , I looked at DeepSeek-R1's code generating capabilities. In this post, I will examine it's ability to answer a more general question. Willy Vlautin Willy Vlautin Photo Larry D. Moore, CC BY 4.0, Wikimedia Commons. Willy Vlautin is my favorite contemporary American author. Vlautin is the author of several novels including The Motel Life, Northline, Lean on Pete, The Free, Don't Skip Out On Me, The Night Always Comes, and The Horse. He is also a songwriter and musician, recording and writing with the bands Richmond Fontaine and The Delines. I have read all of Vlautin's novels. I decided to ask ChatGPT and DeepSeek for authors similar to Willy Vlautin. recommend some authors similar to Willy Vlautin First, I tried ChatGPT o1-min. Here was the response. recommend some authors similar to Willy Vlautin If you appreciate Willy Vlautin's gritty, character-driven narratives that delve into the struggles of everyday life, you might enjoy the follow...

With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.  ~ John von Neumann No telling what you can do with 675B parameters. Generated by Janus Pro  https://www.yeschat.ai/features/janus-pro DeepSeek-R1  has been all over the tech news during the last few weeks, It  scared Wall Street  and forced the  big guys to rethink strategy . DeepSeek-R1 is an open-source AI language model developed by the Chinese startup DeepSeek. There are lots of LLMs available. The big deal with DeepSeek-R1 is that it is open source. The code and parameters are available for download. It was also developed on a shoe string budget compared to the amount spent by outfits like OpenAI for developing their LLMs. I made a few simple tests to compare DeepSeek-R1's performance against OpenAI's o1-mini. o3-mini was not available yet when I made the tests Change Point Detection Code Generation I gave ChatGPT o1-mini and DeepSeek-R1 (using the web ...