-
Notifications
You must be signed in to change notification settings - Fork 8
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Develop/two stage #12
base: master
Are you sure you want to change the base?
Conversation
… and 3) refactor & documentation.
@@ -115,6 +119,7 @@ class vvNet { | |||
|
|||
// vertex vector, context vector, vertex, context, dimension, negative samples, alpha | |||
void UpdateVertex(vector< vector<double> >&, vector< vector<double> >&, long, long, int, int, double); | |||
void UpdateContext(vector< vector<double> >&, vector< vector<double> >&, long, long, int, int, double); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not used in the program?
ICE/ICE.cpp
Outdated
vvnet.UpdatePair(w_vertex, w_context, v1, v2, dim, negative_samples, _alpha); | ||
} | ||
else{ | ||
if (v1 >= vocab_count){ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
if (v1 >= vocab_count)
what's the function of this condition? (you wanna sample a node with an offset vocab_count? if yes, I think we can use "v1+vocab_count" instead)
so now we change the default version to "train stage 1 and then to train stage 2" ? or we'll have an option to do this? |
Unified training method, as oppose to the 2-stage training method, can be achieved via provide the complete ICE network to the 1st stage of the latter. |
Added 1) weighted option: '-w' default=0, 0 uses binary weight and 1 uses real weights
2) Save ET and TT of ICE separately: Allows to save the full ICE network or its ET and TT subnetwork.
3) Refactor & documentation: Put codes into functions and added comments.
Program tested on sample code (unweighted) and self-defined weighted matrix.