 All right, next up is Nathan Caller about a joint paper with Itaiduna. It should be Itaiduna according to our information, but it seems to have changed, also the title has changed. So when this was submitted to us, you can see on this title, it was synchronizing optimally. Apparently in the meantime it changed to how we synchronize efficiently. And then maybe after that it will be how to synchronize practically. Yes, actually we thought that the amendments we do until they certainly will reflect in the program. So both the name of the speaker and the title will change before they certainly will. Okay, anyway, so I'm going to speak about something with that, which is about some simple and nice mathematical header, about some nice mathematics and some application to some nice things. So the integration comes from a morphic section sharing. This was introduced by a late boy, and the late boy was quite shy, he came from 16, as a practical alternative to fully homomorphic encryption. It allows homomorphic immigration and function to be distributed among the parties. We do not interact with each other. They constructed a group based, which is a scheme, but not for all functions, but those described by one particular plan. This was quite successful, but basically more scripted, then quite a few follow-ups in different conferences, many applications. Now, this is for those who understand objectives, now for those who do not, just look at the mathematical problem we have in it. So this was the main problem in the paper of the GI. So the scheme is based on some shared version, which has tables. And since it is repeated many times, then it is very important to reduce the mobility of it. So this amounts to a clear mathematical problem, the following one. We have N random numbers, a region and line. We have two parties. We can adjust and adjust the places on the line, but we do not know who of them is the first and who is the second. Now, which one can query the lines in the most intensive places? They want to see in the comments, that is to choose the same number, but without any communication. So the question is what is the minimum possible error problem? So here is a sequence of the question, which was suggested by the authors. So we have here the sequence of numbers on the line. So each party queries just to take consecutive points, starting with these points for a while, and choosing the minimum. For example, we have T with 5, here both places, and the minimum 11 is the common minimum for both, so they can synchronize. But if we have placed the first two numbers, then they cannot synchronize because one of them chooses the first number and the other chooses the other. Here it is clear that actually the error occurs only if the minimum is in the end point. So the error probability is about 1 with T. The question is whether one can do better. So there are several papers like to improve it, but there will be no asymptotic error rate of 1 with T squared instead of 1 with T. And also if we show a much lower bound, which shows that unless we can't find the script log in the short interval faster than the square root. And so in the interval, we cannot break this model in the square bound. So counting for all standard demographic groups like an entity care center, this is really optimal. At least we believe so. So the techniques are a mixture of a nice mathematical techniques. It starts with some random walks which are with some funny numbers of steps and some funny things. To prove that the random walks work, we use some mathematical theory and to prove the log, we use some discrete free analysis. So some nice mathematical work. Now, applications of that homomorphic scheduling system, this is a random application like a private integration retrieval and other things. So these are placed in our paper in crypto. And now we are continuing to work on this now with the original authors with Elendir and Ruvan about the readings which are outside the topography to some mathematical readers, some brilliant functions and some more cool stuff. So I hope it's nice. We really believe that this algorithm is quite generic and so we'll be using it in other places as well. So, thanks.