|
gptkbp:instance_of
|
gptkb:neural_networks
|
|
gptkbp:application
|
Image Classification
|
|
gptkbp:architecture
|
gptkb:Feedforward_Neural_Network
deep learning
|
|
gptkbp:coat_of_arms
|
gptkb:7
|
|
gptkbp:convolution_kernel_size
|
5x5
|
|
gptkbp:data_preprocessing
|
normalization
|
|
gptkbp:designed_by
|
gptkb:Yann_Le_Cun
|
|
gptkbp:developed_by
|
gptkb:Yann_Le_Cun
|
|
gptkbp:dropout
|
not used
Not used
|
|
gptkbp:evaluates
|
Confusion Matrix
accuracy
|
|
gptkbp:feature_extraction
|
convolutional layers
Learned Features
|
|
gptkbp:fifth_layer
|
fully connected layer
|
|
gptkbp:fifth_layer_type
|
Fully Connected Layer
|
|
gptkbp:first_layer
|
convolutional layer
|
|
gptkbp:first_layer_type
|
Convolutional Layer
|
|
gptkbp:fourth_layer
|
subsampling layer
|
|
gptkbp:fourth_layer_type
|
Subsampling Layer
|
|
gptkbp:framework_used_second
|
gptkb:Tensor_Flow
|
|
gptkbp:framework_used_third
|
gptkb:Py_Torch
|
|
gptkbp:has_website
|
computer vision
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
Le Net-5
|
|
gptkbp:impact
|
gptkb:High
high
|
|
gptkbp:influenced_by
|
gptkb:Neocognitron
backpropagation algorithm
|
|
gptkbp:initialization_method
|
Xavier Initialization
|
|
gptkbp:initiated_by
|
Sigmoid
sigmoid
|
|
gptkbp:input_output
|
32x32 pixels
|
|
gptkbp:inspired_by
|
biological vision systems
|
|
gptkbp:is_a_framework_for
|
gptkb:Caffe
|
|
gptkbp:is_standardized_by
|
not used
Not used
|
|
gptkbp:is_taught_in
|
0.01
|
|
gptkbp:legacy
|
Foundation for modern CNNs
foundation for modern CNNs
|
|
gptkbp:losses
|
Cross-Entropy Loss
cross-entropy loss
|
|
gptkbp:migration
|
not applicable
|
|
gptkbp:momentum
|
0.9
|
|
gptkbp:number_of_filters
|
gptkb:6
16 in third layer
6 in first layer
|
|
gptkbp:number_of_filters_second_layer
|
gptkb:16
|
|
gptkbp:number_of_units_in_fully_connected_layers
|
120
|
|
gptkbp:number_of_units_in_output_layer
|
gptkb:10
|
|
gptkbp:output_activation_function
|
Softmax
softmax
|
|
gptkbp:performance
|
Accuracy
|
|
gptkbp:pooling_type
|
Average Pooling
average pooling
|
|
gptkbp:provides_information_on
|
gptkb:MNIST
not commonly used
Not used
|
|
gptkbp:published_in
|
gptkb:Proceedings_of_the_IEEE
|
|
gptkbp:real_world_application
|
postal code recognition
|
|
gptkbp:real_world_application_second
|
bank check processing
|
|
gptkbp:real_world_application_third
|
character recognition
|
|
gptkbp:regularization
|
L2 Regularization
|
|
gptkbp:related_to
|
gptkb:Artificial_Intelligence
gptkb:machine_learning
gptkb:Deep_Learning
|
|
gptkbp:resolution
|
gptkb:10
10 classes
|
|
gptkbp:second_layer
|
subsampling layer
|
|
gptkbp:second_layer_type
|
Subsampling Layer
|
|
gptkbp:seventh_layer
|
output layer
|
|
gptkbp:seventh_layer_type
|
Output Layer
|
|
gptkbp:sixth_layer
|
fully connected layer
|
|
gptkbp:sixth_layer_type
|
Fully Connected Layer
|
|
gptkbp:subsampling_kernel_size
|
2x2
|
|
gptkbp:successor
|
gptkb:Alex_Net
|
|
gptkbp:test_data_size
|
10,000 images
|
|
gptkbp:third_layer
|
convolutional layer
|
|
gptkbp:third_layer_type
|
Convolutional Layer
|
|
gptkbp:training
|
Backpropagation
varies
|
|
gptkbp:training_data_size
|
60,000 images
|
|
gptkbp:training_epochs
|
60
varies
|
|
gptkbp:tuning
|
stochastic gradient descent
|
|
gptkbp:used_for
|
Handwritten Digit Recognition
handwritten digit recognition
|
|
gptkbp:visualization_tool
|
gptkb:board_game
|
|
gptkbp:year_created
|
gptkb:1998
|
|
gptkbp:year_established
|
gptkb:1998
|