diogo/yap-6.3
Archived
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

ilbfgs

Browse Source
This commit is contained in:
Vitor Santos Costa 2019-03-29 14:37:03 +00:00
parent de153bd479
commit 2af4dae017
3 changed files with 98 additions and 114 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

97
packages/ProbLog/problog/lbdd.yap
View File

@ -58,11 +58,6 @@ update_query(QueryID,Symbol,What_To_Update) :-
) )
). ).
maplist_to_hash([], H0, H0).
maplist_to_hash([I-V|MapList], H0, Hash) :-
rb_insert(H0, V, I, H1),
maplist_to_hash(MapList, H1, Hash).
prob2log(_X,Slope,FactID,V) :- prob2log(_X,Slope,FactID,V) :-
get_fact_probability(FactID, V0), get_fact_probability(FactID, V0),
@ -73,14 +68,11 @@ log2prob(X,Slope,FactID,V) :-
sigmoid(V0, Slope, V). sigmoid(V0, Slope, V).
bind_maplist([], _Slope, _X). bind_maplist([], _Slope, _X).
bind_maplist([Node-Pr|MapList], Slope, X) :- bind_maplist([Node-(Node-Pr)|MapList], Slope, X) :-
Pr <== X[Node], SigPr <== X[Node],
sigmoid(SigPr, Slope, Pr),
bind_maplist(MapList, Slope, X). bind_maplist(MapList, Slope, X).
tree_to_grad([], _, Grad, Grad).
tree_to_grad([Node|Tree], H, Grad0, Grad) :-
node_to_gradient_node(Node, H, GNode),
tree_to_grad(Tree, H, [GNode|Grad0], Grad).
%get_prob(Node, Prob) :- %get_prob(Node, Prob) :-
% query_probability(Node,Prob), !. % query_probability(Node,Prob), !.
@ -97,58 +89,51 @@ gradient(_QueryID, l, _).
gradient(QueryID, g, Slope) :- gradient(QueryID, g, Slope) :-
recorded(QueryID, BDD, _), recorded(QueryID, BDD, _),
query_gradients(BDD,Slope,I,Grad), query_gradients(BDD,Slope,I,Grad),
% writeln(grad(QueryID:I:Grad)),
assert(query_gradient_intern(QueryID,I,p,Grad)), assert(query_gradient_intern(QueryID,I,p,Grad)),
fail. fail.
gradient(QueryID, g, Slope) :- gradient(QueryID, g, Slope) :-
gradient(QueryID, l, Slope). gradient(QueryID, l, Slope).
query_probability( DBDD, Slope, X, Prob) :- query_probabilities( DBDD, Prob) :-
DBDD = bdd(Dir, Tree, MapList), DBDD = bdd(Dir, Tree, _MapList),
bind_maplist(MapList, Slope, X), findall(P, evalp(Tree,P), [Prob0]),
run_sp(Tree, Slope, 1.0, Prob0),
(Dir == 1 -> Prob0 = Prob ; Prob is 1.0-Prob0). (Dir == 1 -> Prob0 = Prob ; Prob is 1.0-Prob0).
evalp( Tree, Prob0) :-
foldl(evalp, Tree, _, Prob0).
query_gradients(bdd(Dir, Tree, MapList),Slope,X,I,Grad) :- query_gradients(bdd(Dir, Tree, MapList),I,IProb,Grad) :-
bind_maplist(MapList, Slope, X), member(I-(_-IProb), MapList),
member(I-_, MapList), % run_grad(Tree, I, Slope, 0.0, Grad0),
run_grad(Tree, I, Slope, 0.0, Grad0), foldl( evalg(I), Tree, _, Grad0),
( Dir = 1 -> Grad = Grad0 ; Grad is -Grad0). ( Dir == 1 -> Grad = Grad0 ; Grad is -Grad0).
evalp( pn(P, _-X, PL, PR), _,P ):-
node_to_gradient_node(pp(P-G,X,L,R), H, gnodep(P,G,X,Id,PL,GL,PR,GR)) :- P is X*PL+ (1.0-X)*(1.0-PR).
rb_lookup(X,Id,H), evalp( pp(P, _-X, PL, PR), _,P ):-
(L == 1 -> GL=0, PL=1 ; L == 0 -> GL = 0, PL=0 ; L = PL-GL), P is X*PL+ (1.0-X)*PR.
(R == 1 -> GR=0, PR=1 ; R == 0 -> GR = 0, PR=0 ; R = PR-GR).
node_to_gradient_node(pn(P-G,X,L,R), H, gnoden(P,G,X,Id,PL,GL,PR,GR)) :- evalg( I, pp(P-G, J-X, L, R), _, G ):-
rb_lookup(X,Id,H), ( number(L) -> PL=L, GL = 0.0 ; L = PL-GL ),
(L == 1 -> GL=0, PL=1 ; L == 0 -> GL = 0, PL=0 ; L = PL-GL), ( number(R) -> PR=R, GR = 0.0 ; R = PR-GR ),
(R == 1 -> GR=0, PR=1 ; R == 0 -> GR = 0, PR=0 ; R = PR-GR). P is X*PL+ (1.0-X)*PR,
(
run_sp([], _, P0, P0). I == J
run_sp(gnodep(P,_G, X, _Id, PL, _GL, PR, _GR).Tree, Slope, _, PF) :- ->
EP = 1.0 / (1.0 + exp(-X * Slope) ), G is X*GL+ (1.0-X)*GR+PL-PR
P is EP*PL+ (1.0-EP)*PR, ;
run_sp(Tree, Slope, P, PF). G is X*GL+ (1.0-X)*GR
run_sp(gnoden(P,_G, X, _Id, PL, _GL, PR, _GR).Tree, Slope, _, PF) :- ).
EP is 1.0 / (1.0 + exp(-X * Slope) ), evalg( I, pn(P-G, J-X, L, R), _,G ):-
P is EP*PL + (1.0-EP)*(1.0 - PR), ( number(L) -> PL=L, GL = 0.0 ; L = PL-GL ),
run_sp(Tree, Slope, P, PF). ( number(R) -> PR=R, GR = 0.0 ; R = PR-GR ),
P is X*PL+ (1.0-X)*(1.0-PR),
run_grad([], _I, _, G0, G0). (
run_grad([gnodep(P,G, X, Id, PL, GL, PR, GR)|Tree], I, Slope, _, GF) :- I == J
EP is 1.0/(1.0 + exp(-X * Slope)), ->
P is EP*PL+ (1.0-EP)*PR, G is X*GL-(1.0-X)*GR+PL-(1-PR)
G0 is EP*GL + (1.0-EP)*GR, ;
% don' t forget the -X G is X*GL- (1.0-X)*GR
( I == Id -> G is G0+(PL-PR)* EP*(1-EP)*Slope ; G = G0 ), ).
run_grad(Tree, I, Slope, G, GF).
run_grad([gnoden(P,G, X, Id, PL, GL, PR, GR)|Tree], I, Slope, _, GF) :-
EP is 1.0 / (1.0 + exp(-X * Slope) ),
P is EP*PL + (1.0-EP)*(1.0 - PR),
G0 is EP*GL - (1.0 - EP) * GR,
( I == Id -> G is G0+(PL+PR-1)*EP*(1-EP)*Slope ; G = G0 ),
run_grad(Tree, I, Slope, G, GF).

24
packages/ProbLog/problog_examples/learn_graph_lbdd.pl
View File

@ -17,7 +17,7 @@
:- use_module('../problog_lbfgs'). :- use_module('../problog_lbfgs').
%% :- if(true). :- if(true).
:- use_module('kbgraph'). :- use_module('kbgraph').
@ -27,9 +27,9 @@
%%%% %%%%
% definition of acyclic path using list of visited nodes % definition of acyclic path using list of visited nodes
%:- else. :- else.
/*
:- set_problog_flag(init_method,(Query,K,Bdd,problog:problog_exact_lbdd(Query,Bdd))). :- Query=path(X,Y), set_problog_flag(init_method,(Query,K,Bdd,problog:problog_exact_lbdd(Query,Bdd))).
path(X,Y) :- path(X,Y,[X],_). path(X,Y) :- path(X,Y,[X],_).
@ -48,8 +48,8 @@ edge(X,Y) :- dir_edge(X,Y).
absent(_,[]). absent(_,[]).
absent(X,[Y|Z]):-X \= Y, absent(X,Z). absent(X,[Y|Z]):-X \= Y, absent(X,Z).
%:- endif. :- endif.
*/
%%%% %%%%
% probabilistic facts % probabilistic facts
% - probability represented by t/1 term means learnable parameter % - probability represented by t/1 term means learnable parameter
@ -84,12 +84,12 @@ example(13,path(4,5),0.57).
example(14,path(4,6),0.51). example(14,path(4,6),0.51).
example(15,path(5,6),0.69). example(15,path(5,6),0.69).
% some examples for learning from proofs: % some examples for learning from proofs:
%example(16,(dir_edge(2,3),dir_edge(2,6),dir_edge(6,5),dir_edge(5,4)),0.032). /*example(16,(dir_edge(2,3),dir_edge(2,6),dir_edge(6,5),dir_edge(5,4)),0.032).
%example(17,(dir_edge(1,6),dir_edge(2,6),dir_edge(2,3),dir_edge(3,4)),0.168). example(17,(dir_edge(1,6),dir_edge(2,6),dir_edge(2,3),dir_edge(3,4)),0.168).
%example(18,(dir_edge(5,3),dir_edge(5,4)),0.14). example(18,(dir_edge(5,3),dir_edge(5,4)),0.14).
%example(19,(dir_edge(2,6),dir_edge(6,5)),0.2). example(19,(dir_edge(2,6),dir_edge(6,5)),0.2).
%example(20,(dir_edge(1,2),dir_edge(2,3),dir_edge(3,4)),0.432). example(20,(dir_edge(1,2),dir_edge(2,3),dir_edge(3,4)),0.432).
*/
%%%%%%%%%%%%%% %%%%%%%%%%%%%%
% test examples of form test_example(ID,Query,DesiredProbability) % test examples of form test_example(ID,Query,DesiredProbability)
% note: ID namespace is shared with training example IDs % note: ID namespace is shared with training example IDs

87
packages/ProbLog/problog_lbfgs.yap
View File

@ -217,7 +217,7 @@
:- yap_flag(unknown,error). :- yap_flag(unknown,error).
% load modules from the YAP library % load modules from the YAP library
:- use_module(library(lists), [member/2,max_list/2, min_list/2, sum_list/2]). :- use_module(library(lists), [member/2,max_list/2, min_list/2, sum_list/2, reverse/2]).
:- use_module(library(system), [file_exists/1, shell/2]). :- use_module(library(system), [file_exists/1, shell/2]).
:- use_module(library(rbtrees)). :- use_module(library(rbtrees)).
:- use_module(library(lbfgs)). :- use_module(library(lbfgs)).
@ -572,20 +572,22 @@ init_one_query(QueryID,Query,_Type) :-
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% if BDD file does not exist, call ProbLog % if BDD file does not exist, call ProbLog
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
fail,
problog_flag(init_method,(Query,N,Bdd,user:graph2bdd(Query,N,Bdd))), problog_flag(init_method,(Query,N,Bdd,user:graph2bdd(Query,N,Bdd))),
!, !,
b_setval(problog_required_keep_ground_ids,false), b_setval(problog_required_keep_ground_ids,false),
(QueryID mod 100 =:= 0 ->writeln(QueryID) ; true), (QueryID mod 100 =:= 0 ->writeln(QueryID) ; true),
Bdd = bdd(Dir, Tree,MapList), Bdd = bdd(Dir, Tree0,MapList),
user:graph2bdd(Query,N,Bdd), user:graph2bdd(Query,N,Bdd),
rb_new(H0), reverse(Tree0,Tree),
maplist_to_hash(MapList, H0, Hash), %rb_new(H0),
tree_to_grad(Tree, Hash, [], Grad), %maplist_to_hash(MapList, H0, Hash),
%tree_to_grad(Tree, Hash, [], Grad),
% ; % ;
% Bdd = bdd(-1,[],[]), % Bdd = bdd(-1,[],[]),
% Grad=[] % Grad=[]
write('.'), write('.'),
recordz(QueryID,bdd(Dir, Grad, MapList),_). recordz(QueryID,bdd(Dir, Tree, MapList),_).
init_one_query(QueryID,Query,_Type) :- init_one_query(QueryID,Query,_Type) :-
% format_learning(3,' ~q example ~q: ~q~n',[Type,QueryID,Query]), % format_learning(3,' ~q example ~q: ~q~n',[Type,QueryID,Query]),
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
@ -594,15 +596,16 @@ init_one_query(QueryID,Query,_Type) :-
b_setval(problog_required_keep_ground_ids,false), b_setval(problog_required_keep_ground_ids,false),
problog_flag(init_method,(Query,_K,Bdd,Call)), problog_flag(init_method,(Query,_K,Bdd,Call)),
!, !,
Bdd = bdd(Dir, Tree, MapList), Bdd = bdd(Dir, Tree0, MapList),
% trace, % trace,
once(Call), once(Call),
rb_new(H0), reverse(Tree0,Tree),
maplist_to_hash(MapList, H0, Hash), %rb_new(H0),
%maplist_to_hash(MapList, H0, Hash),
%Tree \= [], %Tree \= [],
% writeln(Dir:Tree:MapList), % writeln(Dir:Tree:MapList),
tree_to_grad(Tree, Hash, [], Grads), %tree_to_grad(Tree, Hash, [], Grads),
recordz(QueryID,bdd(Dir, Grads, MapList),_). recordz(QueryID,bdd(Dir, Tree, MapList),_).
%======================================================================== %========================================================================
%= %=
@ -780,22 +783,11 @@ inv_sigmoid(T,Slope,InvSig) :-
%= probabilities of the examples have to be recalculated %= probabilities of the examples have to be recalculated
%======================================================================== %========================================================================
:- dynamic index/2.
save_old_probabilities. save_old_probabilities.
mkindex :-
retractall(index(_,_)),
findall(FactID,tunable_fact(FactID,_GroundTruth),L),
foldl(mkindex, L, 0, Count),
assert(count_tunables(Count)).
mkindex(Key,I,I1) :-
I1 is I+1,
assert(index(Key,I),I1).
% vsc: avoid silly search % vsc: avoid silly search
gradient_descent :- gradient_descent :-
mkindex,
problog_flag(sigmoid_slope,Slope), problog_flag(sigmoid_slope,Slope),
% current_iteration(Iteration), % current_iteration(Iteration),
findall(FactID,tunable_fact(FactID,_GroundTruth),L), findall(FactID,tunable_fact(FactID,_GroundTruth),L),
@ -808,8 +800,7 @@ mkindex,
lbfgs_finalize(Solver). lbfgs_finalize(Solver).
set_fact(FactID, Slope, P ) :- set_fact(FactID, Slope, P ) :-
index(FactID, I), X <== P[FactID],
X <== P[I],
sigmoid(X, Slope, Pr), sigmoid(X, Slope, Pr),
(Pr > 0.99 (Pr > 0.99
-> ->
@ -834,16 +825,26 @@ set_tunable(I,Slope,P) :-
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
user:evaluate(LLH_Training_Queries, X,Grad,N,_,_) :- user:evaluate(LLH_Training_Queries, X,Grad,N,_,_) :-
%Handle = user_error, %Handle = user_error,
N1 is N-1,
forall(between(0,N1,I),(Grad[I]<==0.0)),
go( X,Grad, LLs), go( X,Grad, LLs),
sum_list( LLs, LLH_Training_Queries), sum_list( LLs, LLH_Training_Queries).
writeln(LLH_Training_Queries).
test :-
S =.. [f,0-0.9,1-0.8,2-0.6,3-0.7,4-0.5,5-0.4,6-0.7,7-0.2],
functor(S,_,N), N1 is N-1,
problog_flag(sigmoid_slope,Slope),
X <== array[N] of floats,
Grad <== array[N] of floats,
forall(between(0,N1,I),(Grad[I]<==0.0)),
forall(between(1,N,I),(arg(I,S,_-V),inv_sigmoid(V,Slope,V0),I1 is I-1,X[I1]<==V0)),
findall(
LL,
compute_gradient(Grad, X, Slope,LL),
LLs
), sum_list( LLs, LLH_Training_Queries), writeln(LLH_Training_Queries:LLs ),forall(between(0,N1,I),(G<==Grad[I],writeln(I=G))).
update_tunables(X) :-
tunable_fact(FactID,GroundTruth),
set_fact_probability(ID,Prob),
fail.
update_tunables.
go( X,Grad, LLs) :- go( X,Grad, LLs) :-
problog_flag(sigmoid_slope,Slope), problog_flag(sigmoid_slope,Slope),
@ -851,29 +852,27 @@ go( X,Grad, LLs) :-
LL, LL,
compute_gradient(Grad, X, Slope,LL), compute_gradient(Grad, X, Slope,LL),
LLs LLs
),
forall(tunable_fact(FactID,_GroundTruth),
set_fact( FactID, Slope, X)
). ).
compute_gradient( Grad, X, Slope, LL) :- compute_gradient( Grad, X, Slope, LL) :-
user:example(QueryID,_Query,QueryProb), user:example(QueryID,_Query,QueryProb),
recorded(QueryID,BDD,_), recorded(QueryID,BDD,_),
query_probability( BDD, Slope, X, BDDProb), BDD = bdd(_,_,MapList),
bind_maplist(MapList, Slope, X),
query_probabilities( BDD, BDDProb),
LL is (BDDProb-QueryProb)*(BDDProb-QueryProb), LL is (BDDProb-QueryProb)*(BDDProb-QueryProb),
retractall( query_probability_intern( QueryID, _) ), retractall( query_probability_intern( QueryID, _) ),
assert( query_probability_intern( QueryID,BDDProb )), assert( query_probability_intern( QueryID,BDDProb )),
forall( forall(
query_gradients(BDD,Slope,X,I,GradValue), query_gradients(BDD,I,IProb,GradValue),
gradient_pair(BDDProb, QueryProb, Grad, GradValue, Slope, X, I) gradient_pair(BDDProb, QueryProb, Grad, GradValue, I, IProb)
). ).
gradient_pair(BDDProb, QueryProb, Grad, GradValue, Slope, X, I) :- gradient_pair(BDDProb, QueryProb, Grad, GradValue, I, Prob) :-
G0 <== Grad[I], G0 <== Grad[I],
log2prob(X,Slope,I,Prob), GN is G0-GradValue*Prob*(1-Prob)*2*(QueryProb-BDDProb),
%writeln(Prob=BDDProb),
GN is G0+GradValue*BDDProb*(1-BDDProb)*2*(QueryProb-BDDProb),
Grad[I] <== GN. Grad[I] <== GN.
wrap( X, Grad, GradCount) :- wrap( X, Grad, GradCount) :-
@ -890,10 +889,10 @@ wrap( _X, _Grad, _GradCount).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% stop calculate gradient % stop calculate gradient
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
user:progress(FX,_X,_G, _X_Norm,_G_Norm,_Step,_N,CurrentIteration,_Ls,-1) :- user:progress(FX,_X,_G, _X_Norm,_G_Norm,_Step,_N,_CurrentIteration,_Ls,-1) :-
FX < 0, !, FX < 0, !,
format('stopped on bad FX=~4f~n',[FX]). format('stopped on bad FX=~4f~n',[FX]).
user:progress(FX,X,_G,X_Norm,G_Norm,Step,_N, CurrentIteration,Ls,0) :- user:progress(FX,X,_G,X_Norm,G_Norm,Step,_N, Iteration,Ls,0) :-
assertz(current_iteration(Iteration)), assertz(current_iteration(Iteration)),
problog_flag(sigmoid_slope,Slope), problog_flag(sigmoid_slope,Slope),
forall(tunable_fact(FactID,_GroundTruth), set_tunable(FactID,Slope,X)), forall(tunable_fact(FactID,_GroundTruth), set_tunable(FactID,Slope,X)),
@ -901,7 +900,7 @@ user:progress(FX,X,_G,X_Norm,G_Norm,Step,_N, CurrentIteration,Ls,0) :-
save_model, save_model,
X0 <== X[0], sigmoid(X0,Slope,P0), X0 <== X[0], sigmoid(X0,Slope,P0),
X1 <== X[1], sigmoid(X1,Slope,P1), X1 <== X[1], sigmoid(X1,Slope,P1),
format('~d. Iteration : (x0,x1)=(~4f,~4f) f(X)=~4f |X|=~4f |X\'|=~4f Step=~4f Ls=~4f~n',[CurrentIteration,P0 ,P1,FX,X_Norm,G_Norm,Step,Ls]). format('~d. Iteration : (x0,x1)=(~4f,~4f) f(X)=~4f |X|=~4f |X\'|=~4f Step=~4f Ls=~4f~n',[Iteration,P0 ,P1,FX,X_Norm,G_Norm,Step,Ls]).
%======================================================================== %========================================================================