estimation theory – How to detect, track and map a Markov chain

You are receiving a time series whose elements belong to a finite set. Assume the time series is distributed as a Discrete-Time Markov Chain. You receive one element at each time step.

For each time step, your goal is to produce the best possible approximation of the underlying Markov Chain, ideally through a minimal-computation update algorithm.

Has this problem ever been formulated? Does an algorithm for this problem exist? Any little bit of help is very welcome.

Thanks a lot.

P.S.: The update algorithm could include a prediction phase (as Kalman filtering does).

P.S. 2: Bonus points if the solution includes a way to discard the null hypothesis (i.e. to detect it isn’t likely to be a Markov chain, after all).

project management – What is the best way to keep track of versions of multiple micro services delivered as part of a software solution?

We have an enterprise solution which consists of multiple components which are in turn comprised of micro services. We cater to both onprem and hosted models.
The challenge is to keep track of the GIT tag and version details of the components/services that are delivered to customers.
I am looking for a tool that would help track these and open to building one as well. I built a prototype on mvc but looking for something nimbler and easy to host.

Looking for abilities to be able to search for specific version or service. Also the customers might have several labs so a consolidated view will be very helpful.
Like I said, I built a prototype using mvc but would like a more nimble, open source tech stack (if there is no tool already) to maintain the data.
Also, was wondering if the same can be achieved through PowerApps or Sharepoint lists? The grid view/consolidated view along with search are my biggest requirements.

enter image description here

performance – Can we track historically or see queries on server is landing on which schedulers

The ask may sound weird but we are seeing sort of awkward behavior. Or may be we do not have historical data to confirm if it is true.

For our 4 sockets 16 core processor SQL server with total of 64 schedulers ( 4 NUMA but with auto soft numa node we see 8) queries coming in are mostly running ( as seen) on certain numa schedulers. For example numa 4,5,6 has 99% queries running there.

Is there a way in XE or something which tracks queries being run uses which scheduler and we can store this info historically?

screen capture – Can a website track keystrokes like OS or browser related keybind pressing?

There are several ways to track how a user interacts with a website, including mouse position and clicks.

Does the same apply to taking screenshots though? Can a website track browser or system related hotkeys?

For example, which of the following would reveal either clicks or keystrokes (assuming they were pressed while the website had been selected)?

  • Pressing Ctr+Shift+PrintScreen then clicking to select the area I need to take a screenshot of.
  • Ctr+T to open a new tab on Firefox (or any other Firefox keybind).
  • Opening the command line using keybinds.

If I have ‘allow apps to request to track’ on, does the developer still get access to the data listed under app privacy?

The data listed under “App Privacy” is a declaration from the developer to you as the user which types of data they used to track and, and which data they link to you. It is however not a technical or functional specification of features unlocked for you or the developer, and as such do not represent any kind of “blocking” or “system denial” of information made available to the developer.

So this means that for example if the developer has specified that they link “Location” to you, it doesn’t say anything about whether or not iOS makes location information available to the app. You’ll see a separate prompt asking letting you decide whether or not the app will get information from iOS about your location (mainly from the built-in GPS). You can change whether or not the app gets that information from System Settings.

However even if you do not give that permission to the app, the app might still know your location by other means. I.e. for example by extracting location information from your photos, by deriving location information from your IP address, or many other means. The App Privacy declaration is there so that the developer is obligated to inform you whether or not they track you or link you by location by any means.

Note that it is possible for developers to lie or give incorrect information in the App Privacy section. However, if they do, it would certainly give Apple the possibility to remove the app from the App Store if they so choose, and possibly, depending on your jurisdiction, you might have the grounds for a lawsuit.

air travel – Is there a way to track historical averages of flight costs?

Covid-19 has shaken up the flight industry, and I noticed some flights to far off tourist destinations that seem like great deals, e.g. US to Italy for $1000 round trip. If they are so good, I’d like to take advantage while possible; though it would be preferred to wait until Covid-19 is gone to fly. To help with the decision-making process, is there any database of historical flight costs, so I can truly check if these are good prices, or just the usual?

youtube – How can I add a new audio track with uploaded sounds or music?

It seems as if YouTube Studio is still in alpha. There is no obvious way to upload your own sounds or music to use as a track, and you’re stuck with their own canned sounds.

There’s even a link to an “audio library” which allows you to only to download mp3 files from the library of sound effects, but no apparent way to actually upload them (or your own audio files) to the video editor.

Is it possible to upload your own audio files, or is Google still figuring out video editing basics while we’re all waiting impatiently?

Algorithm for associating track ids

Hello Computer Science stackexchange, I am trying to make an object tracking system. I have a module for object detections which feeds into a object tracking module. However, due to objects sometimes not getting detected and a greedy tracking algorithm the output of these two modules end up with 68 objects where the ground truth is only 23 and an illustration of this can be seen in figure 1.

Illustration of positions of every ID in each frame (Y-axis represents IDs and X-axis represents frame number

First I tried to make a greedy algorithm which simply took the closest start frame to an ID’s last frame. This gave some problems because sometimes two IDs would have the same last frame as their closest, which meant the first ID would get the association. I have also tried to introduce euclidean distance as a weight parameter, but the problem is still that I want to find the global best matches (I don’t know the terminology) which can output 23 track ids instead of 68.

So my question is, what field of algorithms is this problem a part of and what do I search?

And how do I solve this problem?

Thanks in advance and sorry if the problem isn’t defined well enough.