Using Orange3 to study Machine Learning (classification & clustering) and Data Analytics (feature analysis). Orange3 is one of Visual Programming Tool for Image analysis and Text mining. About Image analysis, it can provide image embedding such as inception v3, VGG16 and VGG19.
19. 19Manufacturing Technology & Engineering Corning RestrictedDr. Squall Wu 2019
Neural Network
• Neurons per hidden layer: defined as the
ith element represents the number of
neurons in the ith hidden layer.
• E.g. a neural network with 3 layers can be
defined as 2, 3, 2.
• Activation function for the hidden layer:
– Identity: no-op activation, useful to
implement linear bottleneck
– Logistic: the logistic sigmoid function
– tanh: the hyperbolic tan function
– ReLu: the rectified linear unit function
• Solver for weight optimization:
– L-BFGS-B: an optimizer in the family of
quasi-Newton methods
– SGD: stochastic gradient descent
– Adam: stochastic gradient-based
optimizer
• Alpha: L2 penalty (regularization term)
parameter
• Max iterations: maximum number of
iterations
20. 20Manufacturing Technology & Engineering Corning RestrictedDr. Squall Wu 2019
Finding common misclassifications of different models.
23. 24Manufacturing Technology & Engineering Corning RestrictedDr. Squall Wu 2019
Embedders
• InceptionV3 is Google’s deep neural network for image recognition. It
is trained on the ImageNet data set. The model we are using is
available here. For the embedding, Orange3 use the activations of the
penultimate layer of the model, which represents images with vectors.
• SqueezeNet is a deep model for image recognition that achieves
AlexNet-level accuracy on ImageNet with 50x fewer parameters. The
model is trained on the ImageNet dataset. Orange3 re-implemented the
SqueezeNet by using weights from the author’s pretrained model.
Orange3 use activations from pre-softmax (flatten10) layer as an
embedding.
• VGG16 and VGG19 are deep neural networks for image recognition
proposed by Visual Geometry Group from the University of Oxford.
They are trained on the ImageNet data set. Orange3 use a community
implementation of networks with original weights. As an embedding,
Orange3 use activations of the penultimate layer - fc7.