---------------------------------------------------------------------- [Sequential] -------------------------------------------------------------------- [LoweringConvolution] filter size : (3, 3) input shape : {32, 32, 3} output shape : {30, 30, 32} ------------------------------------------------------------------ [ConvolutionIm2Col] input shape : {32, 32, 3} output shape : {3, 3, 3} ------------------------------------------------------------------ [DenseAffine] input shape : {3, 3, 3} output shape : {32} ------------------------------------------------------------------ [ConvolutionCol2Im] input shape : {32} output shape : {30, 30, 32} -------------------------------------------------------------------- [BatchNormalization] input shape : {30, 30, 32} output shape : {30, 30, 32} -------------------------------------------------------------------- [ReLU] input shape : {30, 30, 32} output shape : {30, 30, 32} -------------------------------------------------------------------- [LoweringConvolution] filter size : (3, 3) input shape : {30, 30, 32} output shape : {28, 28, 32} ------------------------------------------------------------------ [ConvolutionIm2Col] input shape : {30, 30, 32} output shape : {3, 3, 32} ------------------------------------------------------------------ [DenseAffine] input shape : {3, 3, 32} output shape : {32} ------------------------------------------------------------------ [ConvolutionCol2Im] input shape : {32} output shape : {28, 28, 32} -------------------------------------------------------------------- [BatchNormalization] input shape : {28, 28, 32} output shape : {28, 28, 32} -------------------------------------------------------------------- [ReLU] input shape : {28, 28, 32} output shape : {28, 28, 32} -------------------------------------------------------------------- [MaxPooling] filter size : (2, 2) input shape : {28, 28, 32} output shape : {14, 14, 32} -------------------------------------------------------------------- [LoweringConvolution] filter size : (3, 3) input shape : {14, 14, 32} output shape : {12, 12, 64} ------------------------------------------------------------------ [ConvolutionIm2Col] input shape : {14, 14, 32} output shape : {3, 3, 32} ------------------------------------------------------------------ [DenseAffine] input shape : {3, 3, 32} output shape : {64} ------------------------------------------------------------------ [ConvolutionCol2Im] input shape : {64} output shape : {12, 12, 64} -------------------------------------------------------------------- [BatchNormalization] input shape : {12, 12, 64} output shape : {12, 12, 64} -------------------------------------------------------------------- [ReLU] input shape : {12, 12, 64} output shape : {12, 12, 64} -------------------------------------------------------------------- [LoweringConvolution] filter size : (3, 3) input shape : {12, 12, 64} output shape : {10, 10, 64} ------------------------------------------------------------------ [ConvolutionIm2Col] input shape : {12, 12, 64} output shape : {3, 3, 64} ------------------------------------------------------------------ [DenseAffine] input shape : {3, 3, 64} output shape : {64} ------------------------------------------------------------------ [ConvolutionCol2Im] input shape : {64} output shape : {10, 10, 64} -------------------------------------------------------------------- [BatchNormalization] input shape : {10, 10, 64} output shape : {10, 10, 64} -------------------------------------------------------------------- [ReLU] input shape : {10, 10, 64} output shape : {10, 10, 64} -------------------------------------------------------------------- [MaxPooling] filter size : (2, 2) input shape : {10, 10, 64} output shape : {5, 5, 64} -------------------------------------------------------------------- [DenseAffine] input shape : {5, 5, 64} output shape : {512} -------------------------------------------------------------------- [BatchNormalization] input shape : {512} output shape : {512} -------------------------------------------------------------------- [ReLU] input shape : {512} output shape : {512} -------------------------------------------------------------------- [DenseAffine] input shape : {512} output shape : {10} ---------------------------------------------------------------------- fitting start : Cifar10DenseCnn 20.45s epoch[ 1] test accuracy : 0.6439 train accuracy : 0.6835 44.52s epoch[ 2] test accuracy : 0.6973 train accuracy : 0.7589 68.48s epoch[ 3] test accuracy : 0.7409 train accuracy : 0.8276 92.54s epoch[ 4] test accuracy : 0.7565 train accuracy : 0.8690 116.80s epoch[ 5] test accuracy : 0.7553 train accuracy : 0.8858 141.12s epoch[ 6] test accuracy : 0.7466 train accuracy : 0.9014 165.58s epoch[ 7] test accuracy : 0.7609 train accuracy : 0.9332 190.38s epoch[ 8] test accuracy : 0.7574 train accuracy : 0.9482 fitting end