DL之CNN：利用自定义DeepConvNet【7+1】算法对mnist数据集训练实现手写数字识别并预测(超过99%)

输出结果

设计思路

核心代码

class DeepConvNet:

   def __init__(self, input_dim=(1, 28, 28),

                conv_param_1 = {'filter_num':16, 'filter_size':3, 'pad':1, 'stride':1},

                conv_param_2 = {'filter_num':16, 'filter_size':3, 'pad':1, 'stride':1},

                conv_param_3 = {'filter_num':32, 'filter_size':3, 'pad':1, 'stride':1},

                conv_param_4 = {'filter_num':32, 'filter_size':3, 'pad':2, 'stride':1},

                conv_param_5 = {'filter_num':64, 'filter_size':3, 'pad':1, 'stride':1},

                conv_param_6 = {'filter_num':64, 'filter_size':3, 'pad':1, 'stride':1},

                hidden_size=50, output_size=10):

       # 初始化权重===========

           ……

   def predict(self, x, train_flg=False):

           ……

   def loss(self, x, t):    

           ……

       return acc / x.shape[0]

   def gradient(self, x, t):

       # forward

       self.loss(x, t)

       # backward

           ……

       return grads

   def save_params(self, file_name="params.pkl"):

           ……

   def load_params(self, file_name="params.pkl"):

           ……

       for i, layer_idx in enumerate((0, 2, 5, 7, 10, 12, 15, 18)):

           self.layers[layer_idx].W = self.params['W' + str(i+1)]

           self.layers[layer_idx].b = self.params['b' + str(i+1)]