See the representation of events

[wl082013]

Hi,
Since the training and test just show the accuracy, they did not provide the learned representations of events. Is there any way to see the learned representations of events?

Another question is about the code: idx_before_bins = x
+ W * y
+ 0
+ W * H * C * p
+ W * H * C * 2 * b Here, what's the intention of doing this? Thanks

[Jee-King]

Hi, have you figured out what this code means now, could you explain it to me? Thanks.

[danielgehrig18]

Here the indices of the voxels that are affected by each event are computed. Since the EST has shape [B x 2 x T x H x W] a general voxel index (in linear form) would be b * W * H * T * 2 + p * W * H * T + t * W * H + y * W + x. Here every event with polarity p and coordinates x,y affects all voxels with varying t. For this reason the index before adding the contribution of t is precomputed. Later the term t * W * H is added (in the for loop) which loops through the different temporal bins.

Hope this helps.

On Tue, Jul 7, 2020 at 3:38 PM JeeKing [email protected] wrote:

Hi, have you figured out what this code means now, could you explain it to me? Thanks.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/uzh-rpg/rpg_event_representation_learning/issues/5#issuecomment-654866916, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBHKKB3M5ALPIQK5GFBUE3R2MQPBANCNFSM4MQTJVZA .

[Jee-King]

Thank you for your reply.

[Andybrizt]

Thank you for your reply.

Jee-King先生您好:

想请问您方便解释作者这些算式的用意吗? idx_before_bins = x + (W * y) + 0 + (W * H * C * p) + (W * H * C * 2 * b)

为何voxel的线性形式会是这样的算式呢?为何index要预先被这样计算呢?

謝謝!