diff --git a/src/main/java/org/dromara/easyai/resnet/ResNetConnectionLine.java b/src/main/java/org/dromara/easyai/resnet/ResNetConnectionLine.java
new file mode 100644
index 0000000..7227711
--- /dev/null
+++ b/src/main/java/org/dromara/easyai/resnet/ResNetConnectionLine.java
@@ -0,0 +1,76 @@
+package org.dromara.easyai.resnet;
+
+import org.dromara.easyai.i.CustomEncoding;
+import org.dromara.easyai.matrixTools.Matrix;
+
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * @author lidapeng
+ * @time 2025/4/11 11:17
+ */
+public class ResNetConnectionLine implements CustomEncoding {
+    private ResBlock lastBlock;//最后一层残差块
+    private int lastSize;//最后一层的特征大小
+    private int allTimes;//隐层神经数量
+    private int nerveSize;//输入神经元数量
+    private int number = 0;
+    private final List<Float> errorValues = new ArrayList<>();
+
+    public void setLastBlock(ResBlock lastBlock, int lastSize, int allTimes, int nerveSize) {
+        this.lastBlock = lastBlock;
+        this.lastSize = lastSize;
+        this.allTimes = allTimes;
+        this.nerveSize = nerveSize;
+    }
+
+    private void addError(Map<Integer, Float> wg) throws Exception {
+        if (wg.size() != nerveSize) {
+            throw new Exception("线性层回传误差数量与预设值不相等");
+        }
+        for (int i = 1; i <= nerveSize; i++) {
+            float error = wg.get(i);
+            if (number == 1) {
+                errorValues.add(error);
+            } else {
+                float value = error + errorValues.get(i - 1);
+                errorValues.set(i - 1, value);
+            }
+        }
+    }
+
+    private void fill(Matrix feature, float value) throws Exception {
+        int x = feature.getX();
+        int y = feature.getY();
+        float myValue = value / (x * y);
+        for (int i = 0; i < x; i++) {
+            for (int j = 0; j < y; j++) {
+                feature.setNub(i, j, myValue);
+            }
+        }
+    }
+
+    private void toError() throws Exception {
+        List<Matrix> errorMatrix = new ArrayList<>();
+        for (Float errorValue : errorValues) {
+            Matrix feature = new Matrix(lastSize, lastSize);
+            float error = errorValue;
+            fill(feature, error);
+            errorMatrix.add(feature);
+        }
+        errorValues.clear();
+        lastBlock.backError(errorMatrix);
+    }
+
+    @Override
+    public void backError(Map<Integer, Float> wg, int id) throws Exception {
+        number++;
+        addError(wg);
+        if (number == allTimes) {
+            number = 0;
+            toError();
+        }
+    }
+}