Merge pull request #106 from bowen-xu/time_projection

Update time projection formula

pynars/Config.py

@@ -81,7 +81,6 @@ class Config:
     n_sequence_attempts = 10
     n_op_condition_attempts = 10
 
-    projection_decay = 0.99
     Td_decision_threshold = 0.51
 
     # what this value represents was originally equal to the termlink record length (10), but we may want to adjust it or make it scaled according to duration since it has more to do with time than # of records.  it can probably be increased several times larger since each item should remain in the recording queue for longer than 1 cycle

pynars/NAL/Functions/TemporalFunctions.py

@@ -6,22 +6,23 @@
 def project(truth: Truth, t_source: int, t_current: int, t_target: int):
     '''
     Reference:
-    [1] OpenNARS 3.1.0 TruthFunctions.java line 492~495:
-        ```
-            public static final float temporalProjection(final long sourceTime, final long targetTime, final long currentTime, Parameters param) {
-                final double a = 100000.0 * param.PROJECTION_DECAY; //projection less strict as we changed in v2.0.0  10000.0 slower decay than 100000.0
-                return 1.0f - abs(sourceTime - targetTime) / (float) (abs(sourceTime - currentTime) + abs(targetTime - currentTime) + a);
-            }
-        ```
-    [2] Hammer, Patrick, Tony Lofthouse, and Pei Wang. "The OpenNARS implementation of the non-axiomatic reasoning system." International conference on artificial general intelligence. Springer, Cham, 2016.
+        p.172 Non-Axiomatic Logic
+            — A Model of Intelligent Reasoning
+            (Second Edition)
+    '''
+    v = abs(t_source - t_target)
 
-        Section 5. Projection and Eternalization
+    t_current_is_in_interval = False
+    if t_source < t_target:
+        if t_current >= t_source and t_current <= t_target: t_current_is_in_interval = True
+    else:
+        if t_current <= t_source and t_current >= t_target: t_current_is_in_interval = True
 
-            $$k_c = \frac{|tB - tT|}{|tB - tC| + |tT - tC|}$$
+    if t_current_is_in_interval: s = 0.5
+    else: s = min(abs(t_source - t_current),abs(t_target-t_current))
 
-            $$c_{new} = (1 - k_c) * c_{old}$$
-    '''
-    c_new = truth.c * (Config.projection_decay ** (t_current - t_source))
+    confidence_discount = 1 - v/(2*s + v)
+    c_new = truth.c * confidence_discount
     return Truth(truth.f, c_new, truth.k)