Timeline for From Bayesian Networks to Neural Networks: how multivariate regression can be transposed to a multi-output network
Current License: CC BY-SA 3.0
36 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 8, 2017 at 5:07 | history | bumped | CommunityBot | This question has answers that may be good or bad; the system has marked it active so that they can be reviewed. | |
| Mar 28, 2017 at 0:50 | comment | added | Tommaso Guerrini | Probably it's better to say I spammed that mailing list .. Thank you very much Luigi by the way.. I'm in that situation where I have no more time to dig into the problems as I should, since I have an incoming deadline.. It seems like STAN is a great tool, but the learning curve is a little steep to really realize its incredible performance (as of now I realized its speed up wrt JAGS) | |
| Mar 28, 2017 at 0:43 | comment | added | lacerbi | Have you tried posing the same question(s) to the Stan users mailing list? They are usually extremely helpful with technical aspects to make a model work. For example, it's possible that issues in your case can be solved with a better parameterization. (Hamiltonian Monte Carlo should mix much faster than Gibbs sampling.) | |
| Mar 28, 2017 at 0:35 | comment | added | Tommaso Guerrini | Last, and then I won't bother you anymore: I don't know much Hamiltonian Monte Carlo, but I have quite an experience with Gibbs sampling.. I was wondering: when increasing computational capacity (I use AWS) should JAGS sampling benefit more from it or STAN sampling or neither ? Sorry for asking a question not in the proper place | |
| Mar 28, 2017 at 0:32 | comment | added | Tommaso Guerrini | To be more specific: the literature does not help much, in the sense that when dealing with supermarket products most methods concentrate just on high selling items which are those who may stock out more easily, generating high costs.. I found no references dealing with all the products | |
| Mar 28, 2017 at 0:30 | comment | added | Tommaso Guerrini | Yes I've tried it, actually I used just STAN,not any other sampling software. For less complex models (for instance not considering mixed effects over brand) and for each product $\lambda_{itjk} = \boldsymbol{\beta} \mathbf{X_t} + \boldsymbol{\eta}_i * \mathbf{Z_{it}}$ it works fine and I'm able to make good inference about what happens.. When increasing complexity I have some convergence problems: I may miss something in my models, but I think the data don't help too since I have lot of skewed predictors, missing data et cetera.. | |
| Mar 28, 2017 at 0:09 | comment | added | lacerbi | Have you tried Stan, or it's not feasible for your problem? Hamiltonian Monte Carlo can be orders of magnitude faster than Gibbs sampling, and scales well to hundreds (or even thousands) of variables. | |
| Mar 27, 2017 at 23:33 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
added 99 characters in body
|
| Mar 24, 2017 at 11:00 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
edited title
|
| Mar 17, 2017 at 11:55 | history | bumped | CommunityBot | This question has answers that may be good or bad; the system has marked it active so that they can be reviewed. | |
| S Feb 17, 2017 at 23:52 | history | bounty ended | CommunityBot | ||
| S Feb 17, 2017 at 23:52 | history | notice removed | CommunityBot | ||
| Feb 12, 2017 at 17:45 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
added 171 characters in body
|
| Feb 12, 2017 at 17:33 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
deleted 2 characters in body
|
| Feb 12, 2017 at 16:19 | answer | added | meh | timeline score: 1 | |
| Feb 12, 2017 at 16:01 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Corrected wrong indices
|
| Feb 12, 2017 at 14:56 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Further explained the bayesian hierarchical model
|
| Feb 11, 2017 at 10:49 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Removed the question about stockouts to highlight the question on neural nets
|
| Feb 11, 2017 at 7:22 | comment | added | Anton Danilov | @Tomasso Guerrini here is possible the answer for you: stats.stackexchange.com/questions/4498/… | |
| Feb 10, 2017 at 20:05 | history | edited | Sycorax♦ | CC BY-SA 3.0 |
added 141 characters in body
|
| Feb 10, 2017 at 19:56 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
highlighted the important part of the question
|
| Feb 10, 2017 at 13:30 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
A possible way of solving the problem
|
| Feb 10, 2017 at 11:38 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
fixed
|
| Feb 10, 2017 at 9:29 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Prioritizing the neural network question
|
| Feb 9, 2017 at 23:06 | history | tweeted | twitter.com/StackStats/status/829828822811938816 | ||
| S Feb 9, 2017 at 22:13 | history | bounty started | Tommaso Guerrini | ||
| S Feb 9, 2017 at 22:13 | history | notice added | Tommaso Guerrini | Canonical answer required | |
| Feb 9, 2017 at 16:12 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
highlighted the important part of the question
|
| Feb 9, 2017 at 15:29 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
added 12 characters in body
|
| Feb 7, 2017 at 17:12 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Further explained the question regarding neural net
|
| Feb 7, 2017 at 15:55 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Further explained the model and why I'm look for those 2 answers
|
| Feb 7, 2017 at 15:19 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
added 11 characters in body
|
| Feb 7, 2017 at 11:55 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
Corrected an index in the second group
|
| Feb 7, 2017 at 11:03 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
added 10 characters in body
|
| Feb 7, 2017 at 9:48 | history | edited | Tommaso Guerrini | CC BY-SA 3.0 |
corrected spelling
|
| Feb 7, 2017 at 9:33 | history | asked | Tommaso Guerrini | CC BY-SA 3.0 |