Merge pull request #47 from NACLab/dev

Nudging over small opt to conv/deconv and arg "x_shape" cleanup (tests passed)

Author: ago109

ngclearn/components/synapses/convolution/convSynapse.py

@@ -19,7 +19,7 @@ class ConvSynapse(JaxComponent): ## base-level convolutional cable
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: 2d shape of input map signal (component currently assumess a square input maps)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -36,15 +36,15 @@ class ConvSynapse(JaxComponent): ## base-level convolutional cable
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((K @ in) * resist_scale) + b
             where `@` denotes convolution
 
         batch_size: batch size dimension of this component
     """
 
     # Define Functions
-    def __init__(self, name, shape, x_size, filter_init=None, bias_init=None, stride=1,
+    def __init__(self, name, shape, x_shape, filter_init=None, bias_init=None, stride=1,
                  padding=None, resist_scale=1., batch_size=1, **kwargs):
         super().__init__(name, **kwargs)
 
@@ -53,6 +53,7 @@ def __init__(self, name, shape, x_size, filter_init=None, bias_init=None, stride
 
         ## Synapse meta-parameters
         self.shape = shape ## shape of synaptic filter tensor
+        x_size, x_size = x_shape
         self.x_size = x_size
         self.Rscale = resist_scale ## post-transformation scale factor
         self.padding = padding
@@ -150,9 +151,12 @@ def help(self): ## component help function
         hyperparams = {
             "shape": "Shape of synaptic filter value matrix; `kernel width` x `kernel height` "
                      "x `number input channels` x `number output channels`",
+            "x_shape": "Shape of any single incoming/input feature map",
             "weight_init": "Initialization conditions for synaptic filter (K) values",
             "bias_init": "Initialization conditions for bias/base-rate (b) values",
-            "resist_scale": "Resistance level output scaling factor (R)"
+            "resist_scale": "Resistance level output scaling factor (R)",
+            "stride": "length / size of stride",
+            "padding": "pre-operator padding to use, i.e., `VALID` `SAME`"
         }
         info = {self.name: properties,
                 "compartments": compartment_props,

ngclearn/components/synapses/convolution/deconvSynapse.py

@@ -19,7 +19,7 @@ class DeconvSynapse(JaxComponent): ## base-level deconvolutional cable
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: 2d shape of input map signal (component currently assumess a square input maps)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -36,15 +36,15 @@ class DeconvSynapse(JaxComponent): ## base-level deconvolutional cable
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((W @.T Rscale) * in) + b
             where `@.T` denotes deconvolution
 
         batch_size: batch size dimension of this component
     """
 
     # Define Functions
-    def __init__(self, name, shape, x_size, filter_init=None, bias_init=None, stride=1,
+    def __init__(self, name, shape, x_shape, filter_init=None, bias_init=None, stride=1,
                  padding=None, resist_scale=1., batch_size=1, **kwargs):
         super().__init__(name, **kwargs)
 
@@ -53,6 +53,7 @@ def __init__(self, name, shape, x_size, filter_init=None, bias_init=None, stride
 
         ## Synapse meta-parameters
         self.shape = shape ## shape of synaptic filter tensor
+        x_size, x_size = x_shape
         self.x_size = x_size
         self.Rscale = resist_scale ## post-transformation scale factor
         self.padding = padding
@@ -138,9 +139,12 @@ def help(self): ## component help function
         hyperparams = {
             "shape": "Shape of synaptic filter value matrix; `kernel width` x `kernel height` "
                      "x `number input channels` x `number output channels`",
+            "x_shape": "Shape of any single incoming/input feature map",
             "weight_init": "Initialization conditions for synaptic filter (K) values",
             "bias_init": "Initialization conditions for bias/base-rate (b) values",
-            "resist_scale": "Resistance level output scaling factor (R)"
+            "resist_scale": "Resistance level output scaling factor (R)",
+            "stride": "length / size of stride",
+            "padding": "pre-operator padding to use, i.e., `VALID` `SAME`"
         }
         info = {self.name: properties,
                 "compartments": compartment_props,

ngclearn/components/synapses/convolution/hebbianConvSynapse.py

@@ -30,7 +30,7 @@ class HebbianConvSynapse(ConvSynapse): ## Hebbian-evolved convolutional cable
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: 2d shape of input map signal (component currently assumess a square input maps)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -49,7 +49,7 @@ class HebbianConvSynapse(ConvSynapse): ## Hebbian-evolved convolutional cable
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((K @ in) * resist_scale) + b
             where `@` denotes convolution
 
@@ -81,11 +81,11 @@ class HebbianConvSynapse(ConvSynapse): ## Hebbian-evolved convolutional cable
     """
 
     # Define Functions
-    def __init__(self, name, shape, x_size, eta=0., filter_init=None, bias_init=None,
+    def __init__(self, name, shape, x_shape, eta=0., filter_init=None, bias_init=None,
                  stride=1, padding=None, resist_scale=1., w_bound=0.,
                  is_nonnegative=False, w_decay=0., sign_value=1., optim_type="sgd",
                  batch_size=1, **kwargs):
-        super().__init__(name, shape, x_size=x_size, filter_init=filter_init,
+        super().__init__(name, shape, x_shape=x_shape, filter_init=filter_init,
                          bias_init=bias_init, resist_scale=resist_scale, stride=stride,
                          padding=padding, batch_size=batch_size, **kwargs)
 
@@ -109,6 +109,15 @@ def __init__(self, name, shape, x_size, eta=0., filter_init=None, bias_init=None
         ## Shape error correction -- do shape correction inference for local updates
         self._init(self.batch_size, self.x_size, self.shape, self.stride,
                    self.padding, self.pad_args, self.weights)
+        self.antiPad = None
+        k_size, k_size, n_in_chan, n_out_chan = self.shape
+        if padding == "SAME":
+            self.antiPad = _conv_same_transpose_padding(self.post.value.shape[1],
+                                                        self.x_size, k_size, stride)
+        elif padding == "VALID":
+            self.antiPad = _conv_valid_transpose_padding(self.post.value.shape[1],
+                                                         self.x_size, k_size, stride)
+
         ########################################################################
 
         ## set up outer optimization compartments
@@ -170,16 +179,16 @@ def evolve(self, opt_params, weights, biases, dWeights, dBiases):
 
     @staticmethod
     def _backtransmit(sign_value, x_size, shape, stride, padding, x_delta_shape,
-                      pre, post, weights): ## action-backpropagating routine
+                      antiPad, post, weights): ## action-backpropagating routine
         ## calc dInputs - adjustment w.r.t. input signal
         k_size, k_size, n_in_chan, n_out_chan = shape
-        antiPad = None
-        if padding == "SAME":
-            antiPad = _conv_same_transpose_padding(post.shape[1], x_size,
-                                                   k_size, stride)
-        elif padding == "VALID":
-            antiPad = _conv_valid_transpose_padding(post.shape[1], x_size,
-                                                    k_size, stride)
+        # antiPad = None
+        # if padding == "SAME":
+        #     antiPad = _conv_same_transpose_padding(post.shape[1], x_size,
+        #                                            k_size, stride)
+        # elif padding == "VALID":
+        #     antiPad = _conv_valid_transpose_padding(post.shape[1], x_size,
+        #                                             k_size, stride)
         dInputs = calc_dX_conv(weights, post, delta_shape=x_delta_shape,
                                stride_size=stride, anti_padding=antiPad)
         ## flip sign of back-transmitted signal (if applicable)
@@ -234,9 +243,12 @@ def help(self): ## component help function
         hyperparams = {
             "shape": "Shape of synaptic filter value matrix; `kernel width` x `kernel height` "
                      "x `number input channels` x `number output channels`",
+            "x_shape": "Shape of any single incoming/input feature map",
             "weight_init": "Initialization conditions for synaptic filter (K) values",
             "bias_init": "Initialization conditions for bias/base-rate (b) values",
             "resist_scale": "Resistance level output scaling factor (R)",
+            "stride": "length / size of stride",
+            "padding": "pre-operator padding to use, i.e., `VALID` `SAME`",
             "is_nonnegative": "Should filters be constrained to be non-negative post-updates?",
             "sign_value": "Scalar `flipping` constant -- changes direction to Hebbian descent if < 0",
             "w_bound": "Soft synaptic bound applied to filters post-update",

ngclearn/components/synapses/convolution/hebbianDeconvSynapse.py

@@ -28,7 +28,7 @@ class HebbianDeconvSynapse(DeconvSynapse): ## Hebbian-evolved deconvolutional ca
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: dimension of input signal (assuming a square input)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -47,7 +47,7 @@ class HebbianDeconvSynapse(DeconvSynapse): ## Hebbian-evolved deconvolutional ca
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((W @.T Rscale) * in) + b
             where `@.T` denotes deconvolution
 
@@ -79,10 +79,10 @@ class HebbianDeconvSynapse(DeconvSynapse): ## Hebbian-evolved deconvolutional ca
     """
 
     # Define Functions
-    def __init__(self, name, shape, x_size, eta=0., filter_init=None, bias_init=None,
+    def __init__(self, name, shape, x_shape, eta=0., filter_init=None, bias_init=None,
                  stride=1, padding=None, resist_scale=1., w_bound=0., is_nonnegative=False,
                  w_decay=0., sign_value=1., optim_type="sgd", batch_size=1, **kwargs):
-        super().__init__(name, shape, x_size=x_size, filter_init=filter_init,
+        super().__init__(name, shape, x_shape=x_shape, filter_init=filter_init,
                          bias_init=bias_init, resist_scale=resist_scale,
                          stride=stride, padding=padding, batch_size=batch_size,
                          **kwargs)
@@ -224,9 +224,12 @@ def help(self): ## component help function
         hyperparams = {
             "shape": "Shape of synaptic filter value matrix; `kernel width` x `kernel height` "
                      "x `number input channels` x `number output channels`",
+            "x_shape": "Shape of any single incoming/input feature map",
             "weight_init": "Initialization conditions for synaptic filter (K) values",
             "bias_init": "Initialization conditions for bias/base-rate (b) values",
             "resist_scale": "Resistance level output scaling factor (R)",
+            "stride": "length / size of stride",
+            "padding": "pre-operator padding to use, i.e., `VALID` `SAME`",
             "is_nonnegative": "Should filters be constrained to be non-negative post-updates?",
             "sign_value": "Scalar `flipping` constant -- changes direction to Hebbian descent if < 0",
             "w_bound": "Soft synaptic bound applied to filters post-update",

ngclearn/components/synapses/convolution/staticConvSynapse.py

@@ -14,7 +14,7 @@ class StaticConvSynapse(ConvSynapse):
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: 2d shape of input map signal (component currently assumess a square input maps)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -31,7 +31,7 @@ class StaticConvSynapse(ConvSynapse):
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((K @ in) * resist_scale) + b
             where `@` denotes convolution
 

ngclearn/components/synapses/convolution/staticDeconvSynapse.py

@@ -14,7 +14,7 @@ class StaticDeconvSynapse(DeconvSynapse):
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: dimension of input signal (assuming a square input)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -31,7 +31,7 @@ class StaticDeconvSynapse(DeconvSynapse):
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((K @ in) * resist_scale) + b
             where `@` denotes convolution
 

ngclearn/components/synapses/convolution/traceSTDPConvSynapse.py

@@ -29,7 +29,7 @@ class TraceSTDPConvSynapse(ConvSynapse): ## trace-based STDP convolutional cable
     Args:
         name: the string name of this cell
 
-        x_size: dimension of input signal (assuming a square input)
+        x_shape: 2d shape of input map signal (component currently assumess a square input maps)
 
         shape: tuple specifying shape of this synaptic cable (usually a 4-tuple
             with number `filter height x filter width x input channels x number output channels`);
@@ -55,7 +55,7 @@ class TraceSTDPConvSynapse(ConvSynapse): ## trace-based STDP convolutional cable
 
         padding: pre-operator padding to use -- "VALID" (none), "SAME"
 
-        resist_scale: aa fixed (resistance) scaling factor to apply to synaptic
+        resist_scale: a fixed (resistance) scaling factor to apply to synaptic
             transform (Default: 1.), i.e., yields: out = ((K @ in) * resist_scale) + b
             where `@` denotes convolution
 
@@ -69,10 +69,10 @@ class TraceSTDPConvSynapse(ConvSynapse): ## trace-based STDP convolutional cable
     """
 
     # Define Functions
-    def __init__(self, name, shape, x_size, A_plus, A_minus, eta=0.,
+    def __init__(self, name, shape, x_shape, A_plus, A_minus, eta=0.,
                  pretrace_target=0., filter_init=None, stride=1, padding=None,
                  resist_scale=1., w_bound=0., w_decay=0., batch_size=1, **kwargs):
-        super().__init__(name, shape, x_size=x_size, filter_init=filter_init,
+        super().__init__(name, shape, x_shape=x_shape, filter_init=filter_init,
                          bias_init=None, resist_scale=resist_scale, stride=stride,
                          padding=padding, batch_size=batch_size, **kwargs)
 
@@ -97,6 +97,15 @@ def __init__(self, name, shape, x_size, A_plus, A_minus, eta=0.,
         ## Shape error correction -- do shape correction inference for local updates
         self._init(self.batch_size, self.x_size, self.shape, self.stride,
                    self.padding, self.pad_args, self.weights)
+        k_size, k_size, n_in_chan, n_out_chan = self.shape
+        if padding == "SAME":
+            self.antiPad = _conv_same_transpose_padding(
+                self.postSpike.value.shape[1],
+                self.x_size, k_size, stride)
+        elif padding == "VALID":
+            self.antiPad = _conv_valid_transpose_padding(
+                self.postSpike.value.shape[1],
+                self.x_size, k_size, stride)
         ########################################################################
 
     def _init(self, batch_size, x_size, shape, stride, padding, pad_args,
@@ -147,17 +156,17 @@ def evolve(self, weights, dWeights):
         self.dWeights.set(dWeights)
 
     @staticmethod
-    def _backtransmit(x_size, shape, stride, padding, x_delta_shape,
-                      preSpike, postSpike, weights): ## action-backpropagating routine
+    def _backtransmit(x_size, shape, stride, padding, x_delta_shape, antiPad,
+                      postSpike, weights): ## action-backpropagating routine
         ## calc dInputs - adjustment w.r.t. input signal
         k_size, k_size, n_in_chan, n_out_chan = shape
-        antiPad = None
-        if padding == "SAME":
-            antiPad = _conv_same_transpose_padding(postSpike.shape[1], x_size,
-                                                   k_size, stride)
-        elif padding == "VALID":
-            antiPad = _conv_valid_transpose_padding(postSpike.shape[1], x_size,
-                                                    k_size, stride)
+        # antiPad = None
+        # if padding == "SAME":
+        #     antiPad = _conv_same_transpose_padding(postSpike.shape[1], x_size,
+        #                                            k_size, stride)
+        # elif padding == "VALID":
+        #     antiPad = _conv_valid_transpose_padding(postSpike.shape[1], x_size,
+        #                                             k_size, stride)
         dInputs = calc_dX_conv(weights, postSpike, delta_shape=x_delta_shape,
                                stride_size=stride, anti_padding=antiPad)
         return dInputs
@@ -213,9 +222,12 @@ def help(self): ## component help function
         hyperparams = {
             "shape": "Shape of synaptic filter value matrix; `kernel width` x `kernel height` "
                      "x `number input channels` x `number output channels`",
+            "x_shape": "Shape of any single incoming/input feature map",
             "weight_init": "Initialization conditions for synaptic filter (K) values",
             "bias_init": "Initialization conditions for bias/base-rate (b) values",
             "resist_scale": "Resistance level output scaling factor (R)",
+            "stride": "length / size of stride",
+            "padding": "pre-operator padding to use, i.e., `VALID` `SAME`",
             "A_plus": "Strength of long-term potentiation (LTP)",
             "A_minus": "Strength of long-term depression (LTD)",
             "eta": "Global learning rate (multiplier beyond A_plus and A_minus)",