def analyzeReceptiveFieldCorrelatedInputs(
inputVectors, sp, params, inputVectors1, inputVectors2):
columnNumber = np.prod(sp.getColumnDimensions())
numInputVector, inputSize = inputVectors.shape
numInputVector1 = params['numInputVectorPerSensor']
numInputVector2 = params['numInputVectorPerSensor']
w = params['numActiveInputBits']
inputSize1 = int(params['inputSize']/2)
inputSize2 = int(params['inputSize']/2)
connectedCounts = np.zeros((columnNumber,), dtype=uintType)
sp.getConnectedCounts(connectedCounts)
numColumns = np.product(sp.getColumnDimensions())
overlapMat1 = np.zeros((numColumns, inputVectors1.shape[0]))
overlapMat2 = np.zeros((numColumns, inputVectors2.shape[0]))
numColumns = np.product(sp.getColumnDimensions())
numInputVector, inputSize = inputVectors.shape
for c in range(numColumns):
connectedSynapses = np.zeros((inputSize,), dtype=uintType)
sp.getConnectedSynapses(c, connectedSynapses)
for i in range(inputVectors1.shape[0]):
overlapMat1[c, i] = percentOverlap(connectedSynapses[:inputSize1],
inputVectors1[i, :inputSize1])
for i in range(inputVectors2.shape[0]):
overlapMat2[c, i] = percentOverlap(connectedSynapses[inputSize1:],
inputVectors2[i, :inputSize2])
sortedOverlapMat1 = np.zeros(overlapMat1.shape)
sortedOverlapMat2 = np.zeros(overlapMat2.shape)
for c in range(numColumns):
sortedOverlapMat1[c, :] = np.sort(overlapMat1[c, :])
sortedOverlapMat2[c, :] = np.sort(overlapMat2[c, :])
fig, ax = plt.subplots(nrows=2, ncols=2)
ax[0, 0].plot(np.mean(sortedOverlapMat1, 0), '-o')
ax[0, 1].plot(np.mean(sortedOverlapMat2, 0), '-o')
fig, ax = plt.subplots(nrows=1, ncols=2)
ax[0].imshow(overlapMat1[:100, :], interpolation="nearest", cmap="magma")
ax[0].set_xlabel('# Input 1')
ax[0].set_ylabel('SP Column #')
ax[1].imshow(overlapMat2[:100, :], interpolation="nearest", cmap="magma")
ax[1].set_xlabel('# Input 2')
ax[1].set_ylabel('SP Column #')
def analyzeReceptiveFieldCorrelatedInputs(
inputVectors, sp, params, inputVectors1, inputVectors2):
columnNumber = np.prod(sp.getColumnDimensions())
numInputVector, inputSize = inputVectors.shape
numInputVector1 = params['numInputVectorPerSensor']
numInputVector2 = params['numInputVectorPerSensor']
w = params['numActiveInputBits']
inputSize1 = int(params['inputSize']/2)
inputSize2 = int(params['inputSize']/2)
connectedCounts = np.zeros((columnNumber,), dtype=uintType)
sp.getConnectedCounts(connectedCounts)
numColumns = np.product(sp.getColumnDimensions())
overlapMat1 = np.zeros((numColumns, inputVectors1.shape[0]))
overlapMat2 = np.zeros((numColumns, inputVectors2.shape[0]))
numColumns = np.product(sp.getColumnDimensions())
numInputVector, inputSize = inputVectors.shape
for c in range(numColumns):
connectedSynapses = np.zeros((inputSize,), dtype=uintType)
sp.getConnectedSynapses(c, connectedSynapses)
for i in range(inputVectors1.shape[0]):
overlapMat1[c, i] = percentOverlap(connectedSynapses[:inputSize1],
inputVectors1[i, :inputSize1])
for i in range(inputVectors2.shape[0]):
overlapMat2[c, i] = percentOverlap(connectedSynapses[inputSize1:],
inputVectors2[i, :inputSize2])
sortedOverlapMat1 = np.zeros(overlapMat1.shape)
sortedOverlapMat2 = np.zeros(overlapMat2.shape)
for c in range(numColumns):
sortedOverlapMat1[c, :] = np.sort(overlapMat1[c, :])
sortedOverlapMat2[c, :] = np.sort(overlapMat2[c, :])
fig, ax = plt.subplots(nrows=2, ncols=2)
ax[0, 0].plot(np.mean(sortedOverlapMat1, 0), '-o')
ax[0, 1].plot(np.mean(sortedOverlapMat2, 0), '-o')
fig, ax = plt.subplots(nrows=1, ncols=2)
ax[0].imshow(overlapMat1[:100, :], interpolation="nearest", cmap="magma")
ax[0].set_xlabel('# Input 1')
ax[0].set_ylabel('SP Column #')
ax[1].imshow(overlapMat2[:100, :], interpolation="nearest", cmap="magma")
ax[1].set_xlabel('# Input 2')
ax[1].set_ylabel('SP Column #')