Paper:
ImageNet classification with deep convolutional neual Network
Achievements:
The model addressed byAlex ETL.achieved top-1 and TOP-5 test error rate of37.5% and17.0%of classifying the 1.2 million high-resolution images in theImageNet LSVRC-2010 ContestInto the different classes.
Model Architecture:
Model Architecture Plot:
contains eight learned layersFive convolutional andthree fully-connected.
The kernels of thesecond, fourth, and fifth convolutional layersis connected-those kernel maps in the previous layer which reside on the same GPU. The kernels of theThird convolutional layer is connected to all kernel maps in the second layer.
response-normalizationlayers Follow theFirst and second convolutional layers.max-pooling Layers,of the kind described in section 3.4,follow both response-normalization layers as well as the fifth convolutional layer. theReLU non-linearityis applied to the output of every convolutional and fully-connected layer.
Interesting Points:
ReLU nonlinearity:speed-up, six times fasterthan an equivalent network with tanh neurons.
overlapping Pooling:enhance accuracy and prevent overfitting, reduces the top-1 and top-5 error rates by 0.4% and 0.3%; training model with overlapping pooling find it slightly m Ore difficult to overfit.
Dropout:prevent overfitting, reduces complex co-adaptations of neurons, since a neuron cannot rely on the Presenc E of particular other neurons. It was, therefore, forced to learn more robust features that was useful in conjunction with many different random subsets O f the other neurons.
