 Hey guys, it's me again Yes, Nathaniel was right. In some sense probabilistic or these of us are common filters But there are some practical differences and by now you might be wondering how hard is it to actually build one of these bad boys yourself And the answer is it's easier than you think because we've done it for you a little bit of history a Good year ago a bunch of people in this group decided to throw everything that we know about probabilistic numerics into one joint software project And this is how problem was born the project. We're still actively working on so if you decide to use it expect some rough edges But at this point of the video feel free to pause click on that link check out the documentation look at the example notebooks But don't forget to come back to this video here and listen to the rest of my story How do you use prop num to solve differential equations if you know to use sci-pi and you kind of know to use problem Let's look at a concrete example here. We're solving the logical terror system of equations We define a right-hand side f we define initial value y0 and If we were to solve this with sci-pi with import solve IVP and call it on these odd e parameters And we'd obtain one of those solution objects and Prop num since we're doing probabilistic numerics were import prop solve IVP and do the same function call again We obtain some sort of solution objects. How do these solution objects compare? Let's look at the result of this little code snippet From this plot, it's clearly evident how superior the problem output is, right? No jerks aside So far they look the same you can evaluate both solutions in between grid points. You can do extrapolation Everything looks identical. However as great as sci-pi is this is where its capabilities stop in prop num We have a full posterior. So for example, we can plot standard deviations or we can even draw joint samples from this posterior And if this is something for you go ahead and check out the prop num code That's a great question No, we can do more than that next to all of the filtering based probabilistic OD solvers as well as the filtering and smoothing Implementations which by the way you can use on the road as well There are other branches of probabilistic numerics represented in prop num. For example, there's probabilistic linear algebra or probabilistic numerical integration So go to the website Check the documentation see whether there is something for you. Let us know what you think and until then enjoy the remaining videos