About confusions in run_placesCNN_unified.py

In this py file, when get `weight_softmax`, there is a statement `weight_softmax[weight_softmax<0] = 0` and I think maybe this is to remove feature maps with negative effects. However, as `weight_softmax[weight_softmax<0] = 0` is implemented, parameters in the network also changes. Is it a good trend?

Start in run_placesCNN_unified.py at the weight_softmax assignment and the subsequent negative-value masking statement. Trace whether this operation shares storage with network parameters, then check the surrounding model and visualization flow. Done means establishing whether parameters are modified and documenting or correcting the behavior accordingly.

Written by the indexing model from the issue text.