Detecting handwritten digits using tensorflow
Initializer - xavier_initializer Optimizer - AdamOptimizer
Nueral Network :
- Batch size = 100
- Layers = 4{ a)hidden_layer_1 (512 neurons) b)hidden_layer_2 (256 neurons) c)hidden_layer_3 (128 neurons) d)output_layer (10 neurons) }
- Optimizer = AdamOptimizer
- Activation function = relu
Weights and biases of each layers are saved as: (W1,W2,W3,WO).csv (B1,B2,B3,BO).csv
Accuracy ~= 98%