.

optim_nn.py

@@ -113,6 +113,70 @@ class LayerNorm(Layer):
 
         return dX
 
+class Denses(Layer): # TODO: rename?
+    # acts as a separate Dense for each row or column. only for 2D arrays.
+
+    def __init__(self, dim, init=init_he_uniform, axis=-1):
+        super().__init__()
+        self.dim = int(dim)
+        self.weight_init = init
+        self.axis = int(axis)
+        self.size = None
+
+    def make_shape(self, shape):
+        super().make_shape(shape)
+        if len(shape) != 2:
+            return False
+
+        assert -len(shape) <= self.axis < len(shape)
+        self.axis = self.axis % len(shape)
+
+        self.output_shape = list(shape)
+        self.output_shape[self.axis] = self.dim
+        self.output_shape = tuple(self.output_shape)
+
+        self.nW = self.dim * np.prod(shape)
+        self.nb = np.prod(self.output_shape)
+        self.size = self.nW + self.nb
+
+        return shape
+
+    def init(self, W, dW):
+        super().init(W, dW)
+
+        ins, outs = np.prod(self.input_shape), np.prod(self.output_shape)
+
+        in_rows = self.input_shape[0]
+        in_cols = self.input_shape[1]
+        out_rows = self.output_shape[0]
+        out_cols = self.output_shape[1]
+
+        self.coeffs = self.W[:self.nW].reshape(in_rows, in_cols, self.dim)
+        self.biases = self.W[self.nW:].reshape(1, out_rows, out_cols)
+        self.dcoeffs = self.dW[:self.nW].reshape(self.coeffs.shape)
+        self.dbiases = self.dW[self.nW:].reshape(self.biases.shape)
+
+        self.coeffs.flat = self.weight_init(self.nW, ins, outs)
+        self.biases.flat = 0
+
+        self.std = np.std(self.W)
+
+    def F(self, X):
+        self.X = X
+        if self.axis == 0:
+            return np.einsum('ixj,xjk->ikj', X, self.coeffs) + self.biases
+        elif self.axis == 1:
+            return np.einsum('ijx,jxk->ijk', X, self.coeffs) + self.biases
+
+    def dF(self, dY):
+        self.dbiases[:] = dY.sum(0, keepdims=True)
+        if self.axis == 0:
+            self.dcoeffs[:] = np.einsum('ixj,ikj->xjk', self.X, dY)
+            return np.einsum('ikj,xjk->ixj', dY, self.coeffs)
+        elif self.axis == 1:
+            self.dcoeffs[:] = np.einsum('ijx,ijk->jxk', self.X, dY)
+            return np.einsum('ijk,jxk->ijx', dY, self.coeffs)
+
 class DenseOneLess(Dense):
     def init(self, W, dW):
         super().init(W, dW)
@@ -122,15 +186,13 @@ class DenseOneLess(Dense):
     def F(self, X):
         np.fill_diagonal(self.coeffs, 0)
         self.X = X
-        Y = X.dot(self.coeffs) + self.biases
-        return Y
+        return X.dot(self.coeffs) + self.biases
 
     def dF(self, dY):
-        dX = dY.dot(self.coeffs.T)
         self.dcoeffs[:] = self.X.T.dot(dY)
         self.dbiases[:] = dY.sum(0, keepdims=True)
         np.fill_diagonal(self.dcoeffs, 0)
-        return dX
+        return dY.dot(self.coeffs.T)
 
 class CosineDense(Dense):
     # paper: https://arxiv.org/abs/1702.05870

optim_nn_core.py

@@ -49,7 +49,7 @@ class Loss:
 class CategoricalCrossentropy(Loss):
     # lifted from theano
 
-    def __init__(self, eps=1e-8):
+    def __init__(self, eps=1e-6):
         self.eps = _f(eps)
 
     def F(self, p, y):
@@ -519,14 +519,20 @@ class Dense(Layer):
 
     def F(self, X):
         self.X = X
-        Y = X.dot(self.coeffs) + self.biases
-        return Y
+        return X.dot(self.coeffs) + self.biases
 
     def dF(self, dY):
-        dX = dY.dot(self.coeffs.T)
+        #Y  = np.einsum('ix,xj->ij',  X,  C)
+        #dX = np.einsum('ix,jx->ij', dY,  C)
+        #dC = np.einsum('xi,xj->ij',  X, dY)
+        # or rather
+        #Y  = np.einsum('ix,xj->ij',  X,  C)
+        #dX = np.einsum('ij,xj->ix', dY,  C)
+        #dC = np.einsum('ix,ij->xj',  X, dY)
+        # that makes sense, just move the pairs around
         self.dcoeffs[:] = self.X.T.dot(dY)
         self.dbiases[:] = dY.sum(0, keepdims=True)
-        return dX
+        return dY.dot(self.coeffs.T)
 
 # Models {{{1