prjn learn flag can't be turned off when the layers and/or the prjn is off about emergent HOT 10 CLOSED

________________________________ From: Xiaonan L. Liu <[email protected]> Sent: Monday, September 30, 2019 10:51:25 PM To: emer/emergent <[email protected]> Cc: emer/emergent <[email protected]>; Xiaonan Liu <[email protected]>; Author <[email protected]> Subject: Re: [emer/emergent] prjn learn flag can't be turned off when the layers and/or the prjn is off (#36) i'm sure it's not just a display only problem as the test results are different. But for the display, if the absolute scale for the color bar is updated immediately after clicking each node (e.g from min -x to min 0), dwt seems display correctly. On Mon, Sep 30, 2019 at 9:51 PM rcoreilly <[email protected]<mailto:[email protected]>> wrote: I just experimented with ra25 and hip examples, and both obey the Off flag, as in the code below. And the IsOff() method checks both the prjn Off flag and both the send and recv layer Off flags. However, it is also definitely the case that when you turn it off, any residual DWt values will still show up in the display. So something strange is happening. we probably need to sit down and debug together. // DWt computes the weight change (learning) -- calls DWt method on sending projections func (ly *Layer) DWt() { for _, p := range ly.SndPrjns { if p.IsOff() { continue } p.(LeabraPrjn).DWt() } } // WtFmDWt updates the weights from delta-weight changes -- on the sending projections func (ly *Layer) WtFmDWt() { for _, p := range ly.SndPrjns { if p.IsOff() { continue } p.(LeabraPrjn).WtFmDWt() } } — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#36?email_source=notifications&email_token=AK6IKTVLMXORNRCYXX4LF7LQMLJNZA5CNFSM4I4AOSN2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD776CNY#issuecomment-536863031>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AK6IKTVEXEKVQSKMJFCKASDQMLJNZANCNFSM4I4AOSNQ>.