09:43:11	 From Martin Hoffmann Petersen : So n is the different classifications?
09:43:44	 From Andreas Malthe Faber : I would guess number of data points we are trying to predict?
09:44:20	 From aagnello : N = how many objects you have
09:44:35	 From aagnello : n = the current object’s index
09:45:31	 From Leon Bozianu : If our estimate of y_n is 1, isn’t the final log term undefined?
09:48:59	 From Kevin Kumar : this is regression right?
09:52:17	 From Camilla : The sampling is called bootstrapping right?
10:02:20	 From Martin Lyskjær Frølund : For the start estimate of y, do we just use a random number?
10:03:44	 From Kasper Hede Nielsen : How does the log loss work in case of multiclass? would the order of the lables have an influence since 1 and 5 is further away than 1 and 2
10:04:06	 From Martin Lyskjær Frølund : Yes
10:05:36	 From aagnello : @Kasper, we’ll cover multi class loss next week, where we re-discuss where the loss functions come from, but it’s actually a good exercise to think about for yourself in the meanwhile ;)
10:10:01	 From Morten Poulsen : When we feed it something we know, I assume it's the same as introducing a bias, is it something we'll cover in the ethics part of the course?
10:10:40	 From Morten Poulsen : ok thanks :)