merged

Author: andrewcropper

README.md

@@ -1,10 +1,8 @@
-# POPPER IS STILL A MAJOR WORK-IN-PROGRESS. BE PREPARED FOR BUGS, DEMONS, AND BACKWARDS-BREAKING CHANGES!
-
 # Popper
 
 Popper is an inductive logic programming (ILP) system.
 If you use Popper for research, please cite the paper [learning programs by learning from failures](https://arxiv.org/abs/2005.02259).
-
+Popper is still a major work-in-progress, so please notify us of bugs or usability issues.
 
 ## Requirements
 
@@ -22,30 +20,50 @@ You can run Popper `python popper.py <input dir>`.
 For instance, running the command `python popper.py examples/dropk` produces the output:
 
 ```prolog
-BEST PROGRAM:
 f(A,B,C):-tail(A,C),one(B)
-f(A,B,C):-f(A,E,D),tail(D,C),decrement(B,E)
-TP: 10, FN: 0, TN: 0+, FP: 0+
+f(A,B,C):-decrement(B,E),f(A,E,D),tail(D,C)
+TP: 10, FN: 0, TN: 10, FP: 0
 ```
 
 Running the command `python popper.py examples/trains` produces the output:
 
 ```prolog
-BEST PROGRAM:
-f(A):-three_wheels(B),long(C),roof_closed(C),has_car(A,B),has_car(A,C)
-TP: 5, FN: 0, TN: 0+, FP: 0+
+f(A):-long(B),roof_closed(B),has_car(A,B),three_wheels(C),has_car(A,C)
+TP: 5, FN: 0, TN: 5, FP: 0
 ```
 
 Take a look at the examples folder for examples.
 
-Note that in the above output, some of the outcomes are `N+`. The reason is that, by default, Popper does not test all the examples during the testing stage, and instead performs *minimal testing*. To test on all the examples, call Popper with the `--test-all` flag.
-
-For instance, running the command `python popper.py examples/trains --test-all` produces the output:
+Popper is an anytime algorithm. To see the intermediate solutions use the `--info` flag. For instance, running the command `python popper.py examples/trains2 --info` produces the output:
 
 ```prolog
-BEST PROGRAM:
-f(A):-long(B),has_car(A,C),roof_closed(B),has_car(A,B),three_wheels(C)
-TP: 5, FN: 0, TN: 5, FP: 0
+NEW BEST PROG 1:
+f(A):-short(B),has_car(A,B)
+TP: 683, FN: 109, TN: 33, FP: 175
+
+NEW BEST PROG 2:
+f(A):-roof_closed(B),has_car(A,B)
+TP: 745, FN: 47, TN: 36, FP: 172
+
+NEW BEST PROG 3:
+f(A):-roof_open(B),has_car(A,B)
+TP: 731, FN: 61, TN: 172, FP: 36
+
+NEW BEST PROG 95:
+f(A):-roof_closed(C),roof_open(B),has_car(A,C),has_car(A,B)
+f(A):-has_load(C,B),has_car(A,C),three_load(B),roof_open(C)
+TP: 721, FN: 71, TN: 208, FP: 0
+
+NEW BEST PROG 382:
+f(A):-has_car(A,C),roof_closed(B),has_car(A,B),roof_open(C)
+f(A):-rectangle(D),has_load(C,B),has_car(A,C),has_load(E,D),triangle(B),has_car(A,E)
+TP: 761, FN: 31, TN: 208, FP: 0
+
+BEST PROG 1271:
+f(A):-roof_closed(C),roof_open(B),has_car(A,C),has_car(A,B)
+f(A):-has_car(A,B),three_load(C),roof_open(B),has_load(B,C)
+f(A):-rectangle(D),has_car(A,B),has_load(B,C),triangle(C),has_load(E,D),has_car(A,E)
+TP: 792, FN: 0, TN: 208, FP: 0
 ```
 
 # Popper settings
@@ -67,3 +85,5 @@ To run with statistics use the flag `--stats`
 To run with a maximum learning time use the flag `--eval-timeout`.
 
 To run in debug mode use the flag `--debug`
+
+To run in information mode use the flag `--debug`

examples/iggp-minimal-decay/bias.pl

@@ -1,19 +1,23 @@
-%% SOLUTION:
-%% next_value(A,B):-does(A,C,D),c_pressButton(D),c5(B),c_player(C)
-%% next_value(A,B):-c_player(E),my_succ(B,C),my_true_value(A,C),does(A,E,D),c_noop(D)
-%% Total programs: 24104
+%% BEST PROGRAM:
+%% next_value(A,B):-c_player(C),c5(B),c_pressButton(D),does(A,C,D)
+%% next_value(A,B):-my_succ(B,D),my_true_value(A,D),c_player(C),c_noop(E),does(A,C,E)
+%% TP: 8, FN: 0, TN: 46, FP: 0
+
+%% Total programs: 2604
 %% Generate:
-%%     Called: 24114 times      Total: 29.08    Mean: 0.001     Max: 0.026
+%%     Called: 2614 times   Total: 2.31     Mean: 0.001     Max: 0.012
 %% Test:
-%%     Called: 24104 times      Total: 18.17    Mean: 0.001     Max: 0.031
+%%     Called: 2604 times   Total: 1.45     Mean: 0.001     Max: 0.040
 %% Build_Rules:
-%%     Called: 24103 times      Total: 4.01     Mean: 0.000     Max: 0.006
+%%     Called: 2603 times   Total: 1.37     Mean: 0.001     Max: 0.007
 %% Ground:
-%%     Called: 24103 times      Total: 0.84     Mean: 0.000     Max: 0.003
+%%     Called: 2603 times   Total: 0.20     Mean: 0.000     Max: 0.003
 %% Add:
-%%     Called: 24103 times      Total: 27.83    Mean: 0.001     Max: 0.006
-%% Total operation time: 79.93s
-%% Total execution time: 80.51s
+%%     Called: 2603 times   Total: 1.56     Mean: 0.001     Max: 0.002
+%% Test_Individual_Rules.Is_Totally_Incomplete:
+%%     Called: 1989 times   Total: 0.10     Mean: 0.000     Max: 0.000
+%% Total operation time: 7.00s
+%% Total execution time: 7.00s
 
 max_clauses(2).
 max_vars(5).
@@ -26,10 +30,10 @@
 body_pred(c_pressButton,1).
 body_pred(c_noop,1).
 body_pred(c_player,1). % comment to make unsat
-body_pred(c1,1). % comment to make easier to solve
-body_pred(c2,1). % comment to make easier to solve
-body_pred(c3,1). % comment to make easier to solve
-body_pred(c4,1). % comment to make easier to solve
+body_pred(c1,1).
+body_pred(c2,1).
+body_pred(c3,1).
+body_pred(c4,1).
 body_pred(c5,1).
 
 type(next_value,(ex,int)).
@@ -45,10 +49,6 @@
 type(c4,(int,)).
 type(c5,(int,)).
 
-functional(my_succ,2).
-irreflexive(my_succ,2).
-functional(my_true_value,2).
-
 %% HACK BECAUSE WE DO NOT LEARN FROM INTERPRETATIONS
 :-
     clause(C),

examples/iggp-rps/bias.pl

@@ -1,34 +1,36 @@
-%% SOLUTION:
-%% next_score(A,B,C):-different(B,D),my_true_score(A,B,C),wins(A,D)
-%% next_score(A,B,C):-my_true_score(A,B,D),wins(A,B),my_succ(D,C)
-%% next_score(A,B,C):-does(A,B,D),different(E,B),my_true_score(A,B,C),does(A,E,D)
-%% Total programs: 8444
+%% BEST PROGRAM:
+%% next_score(A,B,C):-my_true_score(A,B,D),my_succ(D,C),wins(A,B)
+%% next_score(A,B,C):-wins(A,D),my_true_score(A,B,C),different(B,D)
+%% next_score(A,B,C):-does(A,D,E),my_true_score(A,B,C),different(B,D),does(A,B,E)
+%% TP: 108, FN: 0, TN: 356, FP: 0
+
+%% Total programs: 1607
 %% Generate:
-%%     Called: 8456 times   Total: 13.57    Mean: 0.002     Max: 0.246
+%%     Called: 1619 times   Total: 4.64     Mean: 0.003     Max: 0.041
 %% Test:
-%%     Called: 8444 times   Total: 8.76     Mean: 0.001     Max: 0.032
+%%     Called: 1607 times   Total: 7.98     Mean: 0.005     Max: 0.048
 %% Build_Rules:
-%%     Called: 8443 times   Total: 1.37     Mean: 0.000     Max: 0.006
+%%     Called: 1606 times   Total: 3.20     Mean: 0.002     Max: 0.009
 %% Ground:
-%%     Called: 8443 times   Total: 0.67     Mean: 0.000     Max: 0.003
+%%     Called: 1606 times   Total: 0.26     Mean: 0.000     Max: 0.003
 %% Add:
-%%     Called: 8443 times   Total: 9.99     Mean: 0.001     Max: 0.005
-%% Total operation time: 34.38s
-%% Total execution time: 34.70s
+%%     Called: 1606 times   Total: 1.83     Mean: 0.001     Max: 0.004
+%% Test_Individual_Rules.Is_Totally_Incomplete:
+%%     Called: 1539 times   Total: 0.05     Mean: 0.000     Max: 0.001
+%% Total operation time: 17.96s
+%% Total execution time: 18.08s
 
 max_clauses(4).
 max_vars(5).
 max_body(7).
 
 head_pred(next_score,3).
-%% body_pred(draws,2).
 body_pred(wins,2).
-%% body_pred(loses,2).
 body_pred(my_true_score,3).
 body_pred(my_succ,2).
 body_pred(does,3).
 body_pred(beats,2).
-%% body_pred(player,1).
+body_pred(player,1).
 body_pred(different,2).
 
 type(next_score,(ex,player,int)).
@@ -42,11 +44,6 @@
 type(player,(player,)).
 type(different,(player,player)).
 
-functional(beats,2).
-functional(my_succ,2).
-irreflexive(my_succ,2).
-irreflexive(beats,2).
-
 %% HACK BECAUSE WE DO NOT LEARN FROM INTERPRETATIONS
 :-
     clause(C),

examples/trains/bias.pl

@@ -2,9 +2,9 @@
 %% f(A) :- has_car(A,C),long(C),roof_closed(C),has_car(A,B),three_wheels(B).
 %% 0.83s user 0.03s system 99% cpu 0.860 total
 
-max_clauses(4).
+max_clauses(3).
 max_vars(5).
-max_body(5).
+max_body(6).
 
 head_pred(f,1).
 body_pred(has_car,2).
@@ -18,6 +18,7 @@
 body_pred(zero_load,1).
 body_pred(one_load,1).
 body_pred(two_load,1).
+body_pred(three_load,1).
 body_pred(circle,1).
 body_pred(triangle,1).
 body_pred(rectangle,1).
@@ -31,9 +32,10 @@
 type(three_wheels,(car,)).
 type(roof_open,(car,)).
 type(roof_closed,(car,)).
-type(zero_load,(car,)).
-type(one_load,(car,)).
-type(two_load,(car,)).
+type(zero_load,(load,)).
+type(one_load,(load,)).
+type(two_load,(load,)).
+type(three_load,(load,)).
 type(circle,(load,)).
 type(triangle,(load,)).
 type(rectangle,(load,)).
@@ -50,6 +52,7 @@
 direction(zero_load,(in,)).
 direction(one_load,(in,)).
 direction(two_load,(in,)).
+direction(three_load,(in,)).
 direction(circle,(in,)).
 direction(triangle,(in,)).
 direction(rectangle,(in,)).

examples/trains2/bias.pl

@@ -0,0 +1,77 @@
+%% BEST PROGRAM:
+%% f(A):-has_car(A,B),roof_open(C),roof_closed(B),has_car(A,C)
+%% f(A):-three_load(B),roof_open(C),has_car(A,C),has_load(C,B)
+%% f(A):-has_load(C,B),triangle(D),has_car(A,E),rectangle(B),has_load(E,D),has_car(A,C)
+%% TP: 792, FN: 0, TN: 208, FP: 0
+
+%% Total programs: 1274
+%% Generate:
+%%     Called: 1290 times   Total: 11.54    Mean: 0.009     Max: 0.249
+%% Test:
+%%     Called: 1274 times   Total: 14.05    Mean: 0.011     Max: 0.053
+%% Build_Rules:
+%%     Called: 1273 times   Total: 2.29     Mean: 0.002     Max: 0.017
+%% Ground:
+%%     Called: 1273 times   Total: 0.69     Mean: 0.001     Max: 0.006
+%% Add:
+%%     Called: 1273 times   Total: 2.26     Mean: 0.002     Max: 0.004
+%% Test_Individual_Rules.Is_Totally_Incomplete:
+%%     Called: 1259 times   Total: 0.06     Mean: 0.000     Max: 0.002
+%% Total operation time: 30.89s
+%% Total execution time: 31.55s
+
+
+max_clauses(3).
+max_vars(5).
+max_body(6).
+
+head_pred(f,1).
+body_pred(has_car,2).
+body_pred(has_load,2).
+body_pred(short,1).
+body_pred(long,1).
+body_pred(two_wheels,1).
+body_pred(three_wheels,1).
+body_pred(roof_open,1).
+body_pred(roof_closed,1).
+body_pred(zero_load,1).
+body_pred(one_load,1).
+body_pred(two_load,1).
+body_pred(three_load,1).
+body_pred(circle,1).
+body_pred(triangle,1).
+body_pred(rectangle,1).
+
+type(f,(train,)).
+type(has_car,(train,car)).
+type(has_load,(car,load)).
+type(short,(car,)).
+type(long,(car,)).
+type(two_wheels,(car,)).
+type(three_wheels,(car,)).
+type(roof_open,(car,)).
+type(roof_closed,(car,)).
+type(zero_load,(load,)).
+type(one_load,(load,)).
+type(two_load,(load,)).
+type(three_load,(load,)).
+type(circle,(load,)).
+type(triangle,(load,)).
+type(rectangle,(load,)).
+
+direction(f,(in,)).
+direction(has_car,(in,out)).
+direction(has_load,(in,out)).
+direction(short,(in,)).
+direction(long,(in,)).
+direction(two_wheels,(in,)).
+direction(three_wheels,(in,)).
+direction(roof_open,(in,)).
+direction(roof_closed,(in,)).
+direction(zero_load,(in,)).
+direction(one_load,(in,)).
+direction(two_load,(in,)).
+direction(three_load,(in,)).
+direction(circle,(in,)).
+direction(triangle,(in,)).
+direction(rectangle,(in,)).