Computer vison, image processing and optimization
About a year ago, a somewhat sophisticated misinformation campaign was launched in Sweden. The idea of a “brukarkort” was launched in a dozen newly(?) registered websites with names like “dagensprocess,” “brandanalys,” “dokum,” “utrikesgruppen” etc. They all are, in different ways (interviews, news articles, consumer information), spreading the idea that you need a “brukarkort,” a certificate of…
In the post ”Giving GPT-3 a Turing Test,” the model seemingly struggle with nonsensical questions, and the author comments: It also doesn’t know how to say “I don’t know.” […] The lesson here is that if you’re a judge in a Turing test, make sure you ask some nonsense questions, and see if the interviewee […]
If you are following machine learning news, you have likely heard of Open API’s GPT-3 model. They have an API and I thought I’d try to have GPT-3 write some simple code. This it not at all a model meant for coding or anything like that. It is a general language model. In the code […]
In standard Python, you can do this to cache the return value of an expensive function: @lru_cache() def function(n): ... This works well, but it does not work very well for class methods: class MyClass: @lru_cache() def function(self): ... All instances of MyClass will share the same cache. Often, especially for immutable instances, a per-instance […]
In a previous post, I described a system for scheduling shifts to workers that was state of the art. It still is for large problems (small problems can be solved exactly with integer programming). The fixing rule to obtain integer solutions is sometimes called the Ryan-Foster rule [1] and this post will explore how this […]
Nothing in this post is new research; this is all relatively basic machine learning. But it is fun. This is the first day of my vacation. Time for some machine learning! Today, I will create embeddings. I will use the relatively big dataset from the Netflix Prize competition to create embeddings for movies. We develop a way […]
I recently submitted results to the Employee Shift Scheduling public benchmark. My optimizer computed the best known result for all medium-sized and large instances. I currently do not have time to write and publish an article, but a blog could perhaps be a good substitute. My optimizer is based on branch and price. To keep this post short, I will […]
An open problem (as far as I know) in three-dimensional Game of Life is to construct a “glider gun” for a reasonable three-dimensional ruleset. One natural set of rules are 5,6,7/6, in which cells stay alive with 5–7 neighbors and are created with 6 neighbors. In this set of rules, some two-dimensional patterns from the […]
Today, the news in Sweden reported on a study that shows the advantages of delayed cord clamping after 4 years. I thought the results sounded remarkably good, so I looked up the article. Here is a table (from which I have removed most columns): I am not saying this study is wrong, but look at […]
Sameer Agarwal and I just uploaded a paper to arχiv.org: Continuous Optimization for Fields of Experts Denoising Works. We show that simple non-linear least-squares is better than sophisticated discrete optimization methods for certain image denoising problems. I have used these denoising problems as a bechmark and motivation when developing generalized roof duality. But as it turns out, […]