def conv_backward_naive(dout, cache):
"""
A naive implementation of the backward pass for a convolutional layer.
Inputs:
- dout: Upstream derivatives.
- cache: A tuple of (x, w, b, conv_param) as in conv_forward_naive
Returns a tuple of:
- dx: Gradient with respect to x
- dw: Gradient with respect to w
- db: Gradient with respect to b
"""
dx, dw, db = None, None, None
#############################################################################
# TODO: Implement the convolutional backward pass. #
#############################################################################
x, w, b, conv_param, X_col = cache
N, C, H, W = x.shape
F, _, HH, WW = w.shape
_, _, out_h, out_w = dout.shape
stride, pad = conv_param['stride'], conv_param['pad']
db = np.sum(dout, axis=(0, 2, 3))
dout = dout.transpose(1, 0, 2, 3).reshape(F, -1)
dw = np.dot(dout, X_col.T).reshape(w.shape)
dx = np.dot(w.reshape(F, -1).T, dout)
# I then need col2im, and sum up entries that computed multi-times in conv forward
dx.shape = (C, HH, WW, N, out_h, out_w)
from cs231n.im2col_cython import col2im_6d_cython
dx = col2im_6d_cython(dx, N, C, H, W, HH, WW, pad, stride)
#############################################################################
# END OF YOUR CODE #
#############################################################################
return dx, dw, db
def conv_backward_strides(dout, cache):
x, w, b, conv_param, x_cols = cache
stride, pad = conv_param['stride'], conv_param['pad']
N, C, H, W = x.shape
F, _, HH, WW = w.shape
_, _, out_h, out_w = dout.shape
db = np.sum(dout, axis=(0, 2, 3))
# dout_reshaped = dout.transpose(1, 0, 2, 3).reshape(F, -1)
# dw = dout_reshaped.dot(x_cols.T).reshape(w.shape)
# dx_cols = w.reshape(F, -1).T.dot(dout_reshaped)
# dx_cols.shape = (C, HH, WW, N, out_h, out_w)
# dx = col2im_6d_cython(dx_cols, N, C, H, W, HH, WW, pad, stride)
# 省内存
dout = dout.transpose(1, 0, 2, 3).reshape(F, -1)
dw = dout.dot(x_cols.T).reshape(w.shape)
dx_cols = w.reshape(F, -1).T.dot(dout)
dx_cols.shape = (C, HH, WW, N, out_h, out_w)
dx = col2im_6d_cython(dx_cols, N, C, H, W, HH, WW, pad, stride)
return dx, dw, db
def conv_backward_strides(dout, cache): x, w, b, conv_param, x_cols = cache stride, pad = conv_param['stride'], conv_param['pad'] N, C, H, W = x.shape F, _, HH, WW = w.shape _, _, out_h, out_w = dout.shape db = np.sum(dout, axis=(0, 2, 3)) # dout_reshaped = dout.transpose(1, 0, 2, 3).reshape(F, -1) # dw = dout_reshaped.dot(x_cols.T).reshape(w.shape) # dx_cols = w.reshape(F, -1).T.dot(dout_reshaped) # dx_cols.shape = (C, HH, WW, N, out_h, out_w) # dx = col2im_6d_cython(dx_cols, N, C, H, W, HH, WW, pad, stride) # 省内存 dout = dout.transpose(1, 0, 2, 3).reshape(F, -1) dw = dout.dot(x_cols.T).reshape(w.shape) dx_cols = w.reshape(F, -1).T.dot(dout) dx_cols.shape = (C, HH, WW, N, out_h, out_w) dx = col2im_6d_cython(dx_cols, N, C, H, W, HH, WW, pad, stride) return dx, dw, db
def conv_backward_strides(dout, cache):
x, w, b, conv_param, x_cols = cache
stride, pad = conv_param['stride'], conv_param['pad']
N, C, H, W = x.shape
# F, _, HH, WW = w.shape
if (len(w.shape) > 3):
F, _, HH, WW = w.shape
else:
w = w.reshape(1, w.shape[0], w.shape[1], w.shape[2])
F, _, HH, WW = w.shape
_, _, out_h, out_w = dout.shape
db = np.sum(dout, axis=(0, 2, 3))
dout_reshaped = dout.transpose(1, 0, 2, 3).reshape(F, -1)
dw = dout_reshaped.dot(x_cols.T).reshape(w.shape)
dx_cols = w.reshape(F, -1).T.dot(dout_reshaped)
dx_cols.shape = (C, HH, WW, N, out_h, out_w)
dx = col2im_6d_cython(dx_cols, N, C, H, W, HH, WW, pad, stride)
return dx, dw, db
def conv_backward_strides(dout, cache):
x, w, b, conv_param, x_cols = cache
stride, pad = conv_param['stride'], conv_param['pad']
N, C, H, W = x.shape
num_filters, _, filter_height, filter_width = w.shape
_, _, out_h, out_w = dout.shape
db = np.sum(dout, axis=(0, 2, 3))
dout_reshaped = dout.transpose(1, 0, 2, 3).reshape(num_filters, -1)
dw = dout_reshaped.dot(x_cols.T).reshape(w.shape)
dx_cols = w.reshape(num_filters, -1).T.dot(dout_reshaped)
dx_cols.shape = (C, filter_height, filter_width, N, out_h, out_w)
dx = col2im_6d_cython(dx_cols, N, C, H, W, filter_height, filter_width,
pad, stride)
return dx, dw, db
