Conceptor-aided Backpropagation
نویسندگان
چکیده
Catastrophic interference has been a major roadblock in the research of continual learning. Here we propose a variant of the back-propagation algorithm, “conceptor-aided backprop” (CAB), in which gradients are shielded by conceptors against degradation of previously learned tasks. Conceptors have their origin in reservoir computing, where they have been previously shown to overcome catastrophic forgetting. CAB extends these results to deep feedforward networks. On the disjoint and permuted MNIST tasks, CAB outperforms two other methods for coping with catastrophic interference that have recently been proposed.
منابع مشابه
Overcoming Catastrophic Interference by Conceptors
Catastrophic interference has been a major roadblock in the research of continual learning. Here we propose a variant of the back-propagation algorithm, “conceptor-aided back-prop” (CAB), in which gradients are shielded by conceptors against degradation of previously learned tasks. Conceptors have their origin in reservoir computing, where they have been previously shown to overcome catastrophi...
متن کاملClassify Images with Conceptor Network
This article demonstrates a new conceptor network based classifier in classifying images. Mathematical descriptions and analysis are presented. Various tests are experimented in this context using three benchmark datasets: MNIST, CIFAR10 and CIFAR-100. The experiments displayed that conceptor network can offer superior results and flexible configurations than conventional classifiers such as So...
متن کاملSegmentation of Digitized Mammograms Using Self-Organizing Maps in a Breast Cancer Computer Aided Diagnosis System
The objective of this work is to develop a digitized mammograms’ feature extraction approach using Kohonen’s Self-Organizing Maps (SOM). Once developed, the SOM network will be used as the first processing stage in a breast cancer computer aided diagnosis (CAD) system. Its role will be to offer segmented data as input to a second stage dedicated to the diagnosis task, which will be implemented ...
متن کاملSelection Mammogram Texture Descriptors Based on Statistics Properties Backpropagation Structure
Computer Aided Diagnosis (CAD) system has been developed for the early detection of breast cancer, one of the most deadly cancer for women. The benign of mammogram has different texture from malignant. There are fifty mammogram images used in this work which are divided for training and testing. Therefore, the selection of the right texture to determine the level of accuracy of CAD system is im...
متن کاملComupter-aided Diagnosis Tool for the Detection of Cancerous Nodules in X-ray Images
This thesis involves development of a computer-aided diagnosis (CAD) tool for the detection of cancerous nodules in X-ray images. Both cancerous and non-cancerous regions appear with little distinction on an X-ray image. For accurate detection of cancerous nodules, we need to differentiate the cancerous nodules from the noncancerous. We developed an artificial neural network to differentiate th...
متن کامل