Timeline for Best way to pass distribution estimates as a feature into deep learning
Current License: CC BY-SA 4.0
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 17, 2019 at 21:17 | comment | added | JacobJacox | @Acccumulation If u are asking me about correlation between evolution of distributions over time of specific meta model, I honestly didn t examine that... hmm yea so we could check if predictions from.same model but different time step came from static distribution... but that does not rly solve the issue? | |
| Jul 17, 2019 at 16:30 | comment | added | Acccumulation | Is it just a ten piles of 1000 observations each time step, or is there further structure? That is, given two different features or time steps, is there an observation of one that corresponds to an observation of another? | |
| Jul 17, 2019 at 16:23 | history | edited | JacobJacox | CC BY-SA 4.0 |
added 235 characters in body
|
| Jul 17, 2019 at 15:56 | history | became hot network question | |||
| Jul 17, 2019 at 14:24 | vote | accept | JacobJacox | ||
| Jul 17, 2019 at 12:42 | answer | added | user3235916 | timeline score: 1 | |
| Jul 17, 2019 at 10:53 | comment | added | JacobJacox | @StephanKolassa One way for prediction would be to calculate mean and just count how many are >0 and we could sell that as probabilities of classification. But I am sure that is not an optimal solution right... perhaps you can tell more if you understand your distributions better... That's why I want to feed them somehow in a model. | |
| Jul 17, 2019 at 10:53 | comment | added | JacobJacox | @StephanKolassa Yeah sorry. Meta models are very complex by themselves, but basically each operate on specific field (some with text, some with one hot encoded data, some with numeric features) and return possible numeric outputs. Those outputs form distribution of possible changes in target variable. | |
| Jul 17, 2019 at 10:41 | comment | added | Stephan Kolassa | Ah. I think you might get better answers if you could explain a little more what you are trying to do. I have a hard time understanding why you would run ten different meta models that each output 1000 data points, which you then want to combine into a binary classification. Why do you do this? What do the meta models do? Why do you think this pipeline yields better classifications than a simpler model? | |
| Jul 17, 2019 at 10:21 | comment | added | JacobJacox | @user3235916 I have edited the question. | |
| Jul 17, 2019 at 10:18 | history | edited | JacobJacox | CC BY-SA 4.0 |
added 526 characters in body
|
| Jul 17, 2019 at 8:17 | answer | added | Stephan Kolassa | timeline score: 3 | |
| Jul 17, 2019 at 7:57 | comment | added | user3235916 | what are you trying to achieve as an end-goal? | |
| Jul 17, 2019 at 7:47 | history | asked | JacobJacox | CC BY-SA 4.0 |