Deep neural networks via prototype factorization

A system for classifying an image comprising:

• a controller configured to:
• receive a set of images at a deep neural network;
• analyze the images utilizing the deep neural network;
• select an internal layer of the deep neural network that includes one or more activation matrices;
• extract neuron activations at the internal layer utilizing the one or more activation matrices of the internal layer;
• factorize the neuron activations via a matrix factorization algorithm to select prototypes and generate weights for each of the selected prototypes;
• replace the neuron activations of the internal layer with the selected prototypes and the weights for the selected prototypes;
• receive a second set of images at the deep neural network;
• classify the second set of images via the deep neural network using the selected prototypes and the weights for the selected prototypes;
• display the second set of images, the selected prototypes, and the weights for the selected prototypes;
• display predicted results and ground truth for the second set of images analyzed by the deep neural network;
• provide error images based on the predicted results and ground truth;
• identify error prototypes of the selected prototypes associated with the error images;
• rank error weights of the error prototypes of the error images;
• and
• output a new image class based on the error prototypes being one of a top ranked error weights.

A system for classifying a time-series image comprising:

• a controller configured to:
• receive a set of time-series images at the deep neural network;
• analyze the set of time-series images utilizing the Deep Neural Network;
• select an internal layer of the deep neural network that includes one or more activation matrices;
• extract neuron activations at the internal layer utilizing the one or more activation matrices of the internal layer;
• factorize the neuron activations via a matrix factorization algorithm to select prototypes and generate weights for each of the selected prototypes;
• replace the neuron activations of the internal layer with the selected prototypes and the weights for the selected prototypes;
• receive a second set of time-series images at the deep neural network;
• classify the second set of time-series images via the deep neural network using the selected prototypes and the weights for the selected prototypes;
• display the second set of time-series images, the selected prototypes, and the weights for the selected prototypes;
• display predicted results and ground truth for the second set of time-series images analyzed by the deep neural network;
• provide error images based on the predicted results and ground truth;
• identify error prototypes of the selected prototypes associated with the error images;
• rank error weights of the error prototypes of the error images;
• and
• output a new time-series image class based on the error prototypes being one of a top ranked error weights.