You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@madlib.apache.org by nj...@apache.org on 2019/04/24 21:34:37 UTC
[madlib] branch master updated (5a2cea7 -> bd1dda3)
This is an automated email from the ASF dual-hosted git repository.
njayaram pushed a change to branch master
in repository https://gitbox.apache.org/repos/asf/madlib.git.
from 5a2cea7 DL: Add Postgres support
new 8803584 MLP: Fix warm start weights initialization issue
new bd1dda3 MLP: Update error message text in minibatch validation.
The 2 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
src/modules/convex/mlp_igd.cpp | 4 ++--
src/ports/postgres/modules/utilities/minibatch_validation.py_in | 4 ++--
2 files changed, 4 insertions(+), 4 deletions(-)
[madlib] 01/02: MLP: Fix warm start weights initialization issue
Posted by nj...@apache.org.
This is an automated email from the ASF dual-hosted git repository.
njayaram pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/madlib.git
commit 880358453e8b027024f7cca19c5b747d96af6d50
Author: Nandish Jayaram <nj...@apache.org>
AuthorDate: Tue Apr 23 16:56:55 2019 -0700
MLP: Fix warm start weights initialization issue
JIRA: MADLIB-1329
Change code to correctly initialize weights with warm start. In the
C++ layer the weights were not getting initialized correctly from the
previously learnt model, this commit fixes that issue.
Closes #376
---
src/modules/convex/mlp_igd.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/modules/convex/mlp_igd.cpp b/src/modules/convex/mlp_igd.cpp
index 599170b..d838e76 100644
--- a/src/modules/convex/mlp_igd.cpp
+++ b/src/modules/convex/mlp_igd.cpp
@@ -107,7 +107,7 @@ mlp_igd_transition::run(AnyType &args) {
layer_start + j * state.task.model.u[k].rows() + i);
}
}
- layer_start = state.task.model.u[k].rows() * state.task.model.u[k].cols();
+ layer_start += state.task.model.u[k].rows() * state.task.model.u[k].cols();
}
} else {
// initialize the model with appropriate coefficients
@@ -219,7 +219,7 @@ mlp_minibatch_transition::run(AnyType &args) {
layer_start + j * state.model.u[k].rows() + i);
}
}
- layer_start = state.model.u[k].rows() * state.model.u[k].cols();
+ layer_start += state.model.u[k].rows() * state.model.u[k].cols();
}
} else {
// initialize the model with appropriate coefficients
[madlib] 02/02: MLP: Update error message text in minibatch
validation.
Posted by nj...@apache.org.
This is an automated email from the ASF dual-hosted git repository.
njayaram pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/madlib.git
commit bd1dda39b0579ae677f6a1b0b1b2c01457d5c02c
Author: Nandish Jayaram <nj...@apache.org>
AuthorDate: Wed Apr 24 11:12:01 2019 -0700
MLP: Update error message text in minibatch validation.
---
src/ports/postgres/modules/utilities/minibatch_validation.py_in | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/ports/postgres/modules/utilities/minibatch_validation.py_in b/src/ports/postgres/modules/utilities/minibatch_validation.py_in
index 1c92a13..14d97f3 100644
--- a/src/ports/postgres/modules/utilities/minibatch_validation.py_in
+++ b/src/ports/postgres/modules/utilities/minibatch_validation.py_in
@@ -37,6 +37,6 @@ def validate_dependent_var_for_minibatch(table_name, var_name, expr_type=None):
result = plpy.execute(query)
if not result[0]["is_encoded"]:
plpy.error("Dependent variable column {0} in table {1} should be "
- "minibatched and one hot encoded.You might need to re run "
+ "minibatched and one hot encoded. You might need to re run "
"the minibatch_preprocessor function and make sure that "
- "the variable is encoded".format(var_name, table_name))
+ "the variable is encoded.".format(var_name, table_name))