Github user njayaram2 commented on a diff in the pull request:
https://github.com/apache/madlib/pull/272#discussion_r191574947
--- Diff: src/ports/postgres/modules/convex/mlp.sql_in ---
@@ -1474,13 +1480,15 @@ CREATE AGGREGATE MADLIB_SCHEMA.mlp_minibatch_step(
/* warm_start_coeff */ DOUBLE PRECISION[],
/* lambda */ DOUBLE PRECISION,
/* batch_size */ INTEGER,
- /* n_epochs */ INTEGER
+ /* n_epochs */ INTEGER,
+ /* momentum */ DOUBLE PRECISION,
+ /* is_nesterov */ BOOLEAN
)(
STYPE=DOUBLE PRECISION[],
SFUNC=MADLIB_SCHEMA.mlp_minibatch_transition,
m4_ifdef(`__POSTGRESQL__', `',
`prefunc=MADLIB_SCHEMA.mlp_minibatch_merge,')
FINALFUNC=MADLIB_SCHEMA.mlp_minibatch_final,
- INITCOND='{0,0,0,0,0,0,0,0,0,0,0,0}'
+ INITCOND='{0,0,0,0,0,0,0,0,0,0,0,0,0,0}'
);
-------------------------------------------------------------------------
--- End diff --
Can you please update the user docs with momentum and nesterov related
optimizer params? Is it also a good time to update the MLP design doc?
---
