The HP Book Studio G9 is very nearly perfect. I need a computer for data science with which I can also travel far and wide as a digital nomad, and I’ve been searching a while for the best one. Besides programming, movies and writing are the other use cases, and the occasional Civilization game. I […]
HP ZBook Studio 16 Inch G9 Review Read More »
One of the more exciting new features coming soon to AutoTS is the automatic detection of holidays. That’s exactly what it sounds like: figuring out when holidays are occurring and being able to generate that calendar for the future. Firstly, why is this needed? A couple of reasons. The big reason is that holidays are
Holiday Detection for Forecasting Read More »
Reading through documentation for a programming function to find what arguments it likes can be tedious, especially since many of the possible options either don’t work together or aren’t documented. Something I have been including on most of my newer classes is a static method called get_new_params(method=’random’) which does pretty much exactly as it says,
The simple elegance of `.get_new_params()` Read More »
For a model, metrics like MAE and RMSE and so on quantify how accurate a prediction is. However, the vast majority of metrics are symmetric, meaning they care equally about overestimates and underestimates. So far, I have found really one relatively common way for caring about direction: Mean Logarithmic Error (MLE). I also created the
Over and Underestimate Loss Metrics Read More »
It has been impossible in the past to get all three of the largest neural network architectures running in the same Python environment in such a way that they don’t conflict and so that they will also train on GPU. The reason I want all the libraries running in the same environment is that AutoTS
PyTorch, Tensorflow, and MXNet on GPU in the same environment and GPU vs CPU performance Read More »
I was so excited to see the announcement of the M6 forecasting competition. For those who don’t know the M series of competitions (not to be confused with BMW cars) is a long running series of competitions for forecasting time series. It comes from the 1980’s, well predating the modern data science explosion. And this
All the Problems with the M6 Competition (and a few good things) Read More »
Note that a variation on this article may also be posted on the Nousot blog. Time series are a string of numbers varying across time. It could be the weather, sales of tissue paper, electricity demand of a region, or countless other things. There are plenty of useful things to do with this type of
Preprocessing for Time Series Read More »
Ranked Probability Score is an unusual error metric that was designed for meteorology but which has since more often been utilized for assessing models predicting football (soccer) matches. I was forced upon it because of it’s use in the M6 forecasting competition. The purpose of the ranked probability score is to give an accuracy metric
Ranked Probability Score in Python Read More »
Setting up a Python environment is often a pain. Even worse is that sometimes the environment, once it is finally built, is surprisingly slow because the underlying numeric libraries (BLAS, OneAPI, and so on) are improperly configured. This aims to be a fairly definitive guide to fixing those problems as of early 2022. How much
While doing some benchmarking of my AutoTS package I quickly discovered a troubling trend – my newest, fanciest, most expensive Intel i7 CPUs – an Intel 10700 and 1165G7, were performing vastly slower than their older counterparts. I wasn’t terribly surprised to see the 1165G7 having issues: Tiger Lake – a brand new architecture CPU
