v0.4.3 (#32)

* log_p and log_x in entropy_pooling loop. (#30) * v0.4.3rc (#31) * Poetry update * Docs update Co-authored-by: ft-cama <[email protected]>
fortitudo-tech · May 25, 2022 · 2651fcf · 2651fcf
1 parent 654548d
commit 2651fcf
Show file tree

Hide file tree

Showing 14 changed files with 107 additions and 102 deletions.
diff --git a/docs/build/doctrees/environment.pickle b/docs/build/doctrees/environment.pickle
diff --git a/docs/build/html/contributing.html b/docs/build/html/contributing.html
@@ -31,7 +31,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/documentation.html b/docs/build/html/documentation.html
@@ -32,7 +32,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/examples.html b/docs/build/html/examples.html
@@ -31,7 +31,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/genindex.html b/docs/build/html/genindex.html
@@ -28,7 +28,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/index.html b/docs/build/html/index.html
@@ -30,7 +30,7 @@
             <a href="#" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/objects.inv b/docs/build/html/objects.inv
diff --git a/docs/build/html/py-modindex.html b/docs/build/html/py-modindex.html
@@ -31,7 +31,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/references.html b/docs/build/html/references.html
@@ -30,7 +30,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">

diff --git a/docs/build/html/search.html b/docs/build/html/search.html
@@ -31,7 +31,7 @@
             <a href="index.html" class="icon icon-home"> fortitudo.tech
           </a>
               <div class="version">
-                v0.4.2
+                v0.4.3
               </div>
 <div role="search">
   <form id="rtd-search-form" class="wy-form" action="#" method="get">

diff --git a/docs/source/conf.py b/docs/source/conf.py
@@ -38,7 +38,7 @@
 author = 'Fortitudo Technologies'
 
 # The full version, including alpha/beta/rc tags
-version = 'v0.4.2'
+version = 'v0.4.3'
 
 
 # -- General configuration ---------------------------------------------------

diff --git a/fortitudo/tech/entropy_pooling.py b/fortitudo/tech/entropy_pooling.py
@@ -34,81 +34,84 @@ def entropy_pooling(
     Returns:
         Posterior probability vector with shape (S, 1).
     """
+    log_p = np.log(p)
     if G is None:
         solution = minimize(
-            _dual_equality, x0=np.zeros(A.shape[0]), args=(p, A, b),
+            _dual_equality, x0=np.zeros(A.shape[0]), args=(log_p, A, b),
             method='Newton-CG', jac=True, hess=_hessian_equality,
             options={'maxiter': 10000})
-        q = np.exp(np.log(p) - 1 - A.T @ solution.x[:, np.newaxis])
+        q = np.exp(log_p - 1 - A.T @ solution.x[:, np.newaxis])
     else:
         len_b = len(b)
         len_h = len(h)
         len_bh = len_b + len_h
         bounds = Bounds([-np.inf] * len_b + [0] * len_h, [np.inf] * len_bh)
         lhs = np.vstack((A, G))
         solution = minimize(
-            _dual_inequality, x0=np.zeros(len_bh), args=(p, lhs, np.vstack((b, h))),
+            _dual_inequality, x0=np.zeros(len_bh), args=(log_p, lhs, np.vstack((b, h))),
             method='TNC', jac=True, bounds=bounds, options={'maxiter': 10000})
-        q = np.exp(np.log(p) - 1 - lhs.T @ solution.x[:, np.newaxis])
+        q = np.exp(log_p - 1 - lhs.T @ solution.x[:, np.newaxis])
     return q
 
 
 def _dual_equality(
-        equality_multipliers: np.ndarray, p: np.ndarray, A: np.ndarray,
+        equality_multipliers: np.ndarray, log_p: np.ndarray, A: np.ndarray,
         b: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
     """Function computing equality constrained objective and gradient.
 
     Args:
         equality_multipliers: Lagrange multipliers with shape (M,).
-        p: Prior probability vector with shape (S, 1).
+        log_p: Log of prior probability vector with shape (S, 1).
         A: Equality constraint matrix with shape (M, S).
         b: Equality constraint vector with shape (M, 1).
 
     Returns:
         Tuple containing the dual objective value and gradient.
     """
     equality_multipliers = equality_multipliers[:, np.newaxis]
-    x = np.exp(np.log(p) - 1 - A.T @ equality_multipliers)
+    log_x = log_p - 1 - A.T @ equality_multipliers
+    x = np.exp(log_x)
     gradient = b - A @ x
-    dual_objective = x.T @ (np.log(x) - np.log(p)) - equality_multipliers.T @ gradient
+    dual_objective = x.T @ (log_x - log_p) - equality_multipliers.T @ gradient
     return -dual_objective, gradient.flatten()
 
 
 def _dual_inequality(
-        lagrange_multipliers: np.ndarray, p: np.ndarray, lhs: np.ndarray,
+        lagrange_multipliers: np.ndarray, log_p: np.ndarray, lhs: np.ndarray,
         rhs: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
     """Function computing inequality constrained objective and gradient.
 
     Args:
-        lagrange_multipliers: Lagrange multipliers with shape (N + M,).
-        p: Prior probability vector with shape (S, 1).
-        lhs: Matrix with shape (N + M, S).
-        rhs: Vector with shape (N + M, 1).
+        lagrange_multipliers: Lagrange multipliers with shape (M + N,).
+        log_p: Log of prior probability vector with shape (S, 1).
+        lhs: Matrix with shape (M + N, S).
+        rhs: Vector with shape (M + N, 1).
 
     Returns:
         Tuple containing the dual objective value and gradient.
     """
     lagrange_multipliers = lagrange_multipliers[:, np.newaxis]
-    x = np.exp(np.log(p) - 1 - lhs.T @ lagrange_multipliers)
+    log_x = log_p - 1 - lhs.T @ lagrange_multipliers
+    x = np.exp(log_x)
     gradient = rhs - lhs @ x
-    dual_objective = x.T @ (np.log(x) - np.log(p)) - lagrange_multipliers.T @ gradient
+    dual_objective = x.T @ (log_x - log_p) - lagrange_multipliers.T @ gradient
     return -1000 * dual_objective, 1000 * gradient
 
 
 def _hessian_equality(
-        equality_multipliers: np.ndarray, p: np.ndarray,
+        equality_multipliers: np.ndarray, log_p: np.ndarray,
         A: np.ndarray, b: np.ndarray) -> np.ndarray:
     """Function computing equality constrained hessian.
 
     Args:
         equality_multipliers: Lagrange multipliers with shape (M,).
-        p: Prior probability vector with shape (S, 1).
+        log_p: Log of prior probability vector with shape (S, 1).
         A: Equality constraint matrix with shape (M, S).
         b: Equality constraint vector with shape (M, 1).
 
     Returns:
         Hessian matrix with shape (M, M).
     """
-    x = np.exp(np.log(p) - 1 - A.T @ equality_multipliers[:, np.newaxis])
+    x = np.exp(log_p - 1 - A.T @ equality_multipliers[:, np.newaxis])
     hessian = A @ (x @ np.ones((1, A.shape[0])) * A.T)
     return hessian