Ira Massimo, Dwayne, Sarah discussed on Data Skeptic

Data Skeptic
|

Automatic TRANSCRIPT

It really works. And it's not because this moment we can just apply this sarah model in different cases for example if the time series than the violent coherent correlation for this to series we gave some high zorich co lines which means we generally the model based on the ground. Truth more as But in billboard problems as i eyesight this well currently counts as part true dwayne up some more specific data out driving spectral. Honesty's was you said you may not have you know a lot real world data sets but we have some toy examples that play out the method very well in the paper. Could you describe maybe the circumstances under which you can really see the performance difference between the arniston and the traditional zarema is zinc the second section of our paper. We give some a toy examples to illustrate the advantage of for example. If we investigate. Sarah i'm on my seeing more do based on a real. I'm i'm is four one and we stitch the you can see the paper the finger and we have Used the one hundred times of motorcars munition. An average prediction messy. And we'll find onto the era of ira massimo these much smaller than there on model this aging of this. The days i say hoa generate betas follows standard procedure with spokes. Jason message estimate says no. I remind which is tom order with a standard procedure which can be employed in our pancakes and the appreciable used to traveling says no difference but if we use ama- seymour do we use our procedure direction. And the week we get better results. How would you describe the improvement. Is it incremental or like an order of magnitude improvement as and this momentum is incremental. And do you have sense of any real world. Problems may be specific industries or areas of science where this technique would likely be especially beneficial. I think we're going to apply this leukaemia problems in retail because even focused on retail is i. He's very important per second. This problem is very complex. For example if you want to predict the demand of product category which means for example if you want to protect the demand although phoolan all therefore or cities is lower than difficult because this anger neighbor were high for example if you want to predict the demand for specific product as for example want predicts the that the amount of the newest iphone and the air is our safety and in this specific aero maybe is very noisy. something's and also is the time energy state. Say if you want to predict the the demand of iphone for each day to pacific area this problem if you candidate. Some statisticians may seagate's let's cover is unpredictable. Which means this cobra. How many noise actually from the practice point you. If you'll come predict this so-called very noise data you can solve their high on aggregate problems. Where will be because you is up to the batteries us so this is a sin. But you you kind of worry. Aggregate enable the state has is you. Oh you apply the traditional mazars. I remarked remind lord and easy. So we're going to our new candidate. Tom with guests that maybe we can use their. I remind seymour and this moment because some results were going to publish Is not appropriate to discuss this unpublished work. And.

Coming up next