Timeline for Evaluate performance of non-negative matrix factorization (NMF)
Current License: CC BY-SA 3.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 29, 2016 at 16:39 | comment | added | Andrew | I have never used NMF with missing data. This post showed up when I searched. The nmf_update.lsnmf function seems to allow weights, which you can set to 0 for missing entries. | |
| Mar 26, 2016 at 22:26 | comment | added | amoeba | @Andrew: What software/library/function are you using to perform NMF in the presence of missing values? Is it straightforward? | |
| Mar 26, 2016 at 22:21 | history | edited | Andrew | CC BY-SA 3.0 |
added 3 characters in body
|
| Mar 26, 2016 at 22:21 | comment | added | Andrew | The percent to set missing is pretty subjective. I usually do around 20% or 25%. If you set too little missing, the validation will be noisy. If you set too much missing, you may not be training the model to its full capacity. You could also do this repeatedly or do cross validation. | |
| Mar 25, 2016 at 15:05 | comment | added | amoeba | +1. My answer here stats.stackexchange.com/questions/111205 is related to your third paragraph (cc to @pir). | |
| Mar 25, 2016 at 14:58 | vote | accept | pir | ||
| Mar 25, 2016 at 14:58 | comment | added | pir | Interesting. Hadn't thought about the last option for evaluating NMF. How many counts would you set missing: 1%, 5%, 10%? And just at random throughout the matrix? | |
| Mar 25, 2016 at 14:41 | history | answered | Andrew | CC BY-SA 3.0 |