Over the past decade, deep learning has made a name for itself by producing state-of-the-heart results across computer vision, natural language processing, speech recognition, and many more such applications. Some of the models that human researchers have designed and engineered have also gained popularity, including AlexNet, Inception, VGGNet, ResNet, and DenseNet; some of them are now the go-to standard for their respective tasks. However, it seems that the better the model gets, the more complex the architecture becomes, especially with the introduction of residual connections between convolutional layers. The task of designing a high-performance neural network has thus become a very arduous one. Hence the question arises: is it possible for an algorithm to learn how to generate neural network architectures?
As the title of this...
