proof-of-portfolio 0.0.103__tar.gz → 0.0.105__tar.gz

{proof_of_portfolio-0.0.103 → proof_of_portfolio-0.0.105}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: proof-of-portfolio
-Version: 0.0.103
+Version: 0.0.105
 Summary: Zero-Knowledge Proof framework for verifiable, private portfolio performance metrics
 Author-email: "Inference Labs, Inc." <info@inferencelabs.com>
 Requires-Python: >=3.10

{proof_of_portfolio-0.0.103 → proof_of_portfolio-0.0.105}/proof_of_portfolio/_version.py

@@ -1,2 +1,2 @@
 # This file is auto-generated during build
-__version__ = "0.0.103"
+__version__ = "0.0.105"

{proof_of_portfolio-0.0.103 → proof_of_portfolio-0.0.105}/proof_of_portfolio/circuits/components/src/core/omega.nr

@@ -18,8 +18,8 @@ pub fn omega(
         if use_weighting {
             let mut product_sum_positive: i64 = 0;
             let mut product_sum_negative: i64 = 0;
-            let mut sum_weights_positive: i64 = 0;
-            let mut sum_weights_negative: i64 = 0;
+            let mut sum_weights_positive_raw: i64 = 0;
+            let mut sum_weights_negative_raw: i64 = 0;
 
             for i in 0..ARRAY_SIZE {
                 if (i as u32) < actual_len {
@@ -27,31 +27,44 @@ pub fn omega(
                     let log_return = log_returns[i];
                     if log_return > 0 {
                         product_sum_positive += log_return * weight;
-                        sum_weights_positive += weight;
+                        sum_weights_positive_raw += weight;
                     } else {
                         product_sum_negative += log_return * weight;
-                        sum_weights_negative += weight;
+                        sum_weights_negative_raw += weight;
                     }
                 }
             }
 
-            let mean_pos = if sum_weights_positive != 0 {
-                product_sum_positive / sum_weights_positive
+            // Apply max with omega_loss_min like in Python reference
+            let sum_weights_positive = if sum_weights_positive_raw >= omega_loss_min {
+                sum_weights_positive_raw
             } else {
-                0
+                omega_loss_min
             };
-            let mean_neg = if sum_weights_negative != 0 {
-                (-product_sum_negative) / sum_weights_negative
+            let sum_weights_negative = if sum_weights_negative_raw >= omega_loss_min {
+                sum_weights_negative_raw
             } else {
-                0
+                omega_loss_min
             };
 
-            let effective_denominator = if mean_neg >= omega_loss_min {
-                mean_neg
+            // Apply cross-multiplication with scaling to prevent overflow
+            // Scale down both factors before multiplication, then scale back up
+            let scale_down = SCALE / 1000; // Scale down by 1000 to prevent overflow
+            let product_sum_positive_scaled = product_sum_positive / scale_down;
+            let product_sum_negative_scaled = product_sum_negative / scale_down;
+
+            let positive_sum_weighted = product_sum_positive_scaled * sum_weights_negative;
+            let negative_sum_weighted = product_sum_negative_scaled * sum_weights_positive;
+
+            let effective_denominator = if (-negative_sum_weighted) >= (omega_loss_min / scale_down) {
+                -negative_sum_weighted
             } else {
-                omega_loss_min
+                omega_loss_min / scale_down
             };
-            (mean_pos * RATIO_SCALE_FACTOR) / effective_denominator
+
+            // Apply scale_down correction to RATIO_SCALE_FACTOR
+            let adjusted_ratio_scale = RATIO_SCALE_FACTOR * scale_down;
+            (positive_sum_weighted * adjusted_ratio_scale) / effective_denominator
         } else {
             let mut positive_sum: i64 = 0;
             let mut negative_sum: i64 = 0;