This paper introduces the fast adaptive stochastic function accelerator (FASFA) for gradient-based optimization of stochastic objective functions. It works based on Nesterov-enhanced first and second momentum estimates. The method is simple and effective during implementation because it has intuitive/familiar hyperparameterization. The training dynamics can be progressive or conservative depending on the decay rate sum. It works well with a low learning rate and mini batch size. Experiments and statistics showed convincing evidence that FASFA could be an ideal candidate for optimizing stochastic objective functions, particularly those generated by multilayer perceptrons with convolution and dropout layers. In addition, the convergence properties and regret bound provide results aligning with the online convex optimization framework. In a first of its kind, FASFA addresses the growing need for diverse optimizers by pro-viding next-generation training dynamics for artificial intelligence algorithms. Future experiments could modify FASFA based on the infinity norm. FASFA: A Novel Next-Generation Backpropagation Optimizer

Deep-learning models estimate values using backpropagation. The activation function within hidden layers is a critical component to minimizing loss in deep neural-networks. Rectified Linear (ReLU) has been the dominant activation function for the past decade. Swish and Mish are newer activation functions that have shown to yield better results than ReLU given specific circumstances. Phish is a novel non-monotonic activation function proposed here. It is a composite function defined as f(x) = xTanH(GELU(x)), where no discontinuities are apparent in the differentiated graph on the domain observed. Four generalized networks were constructed using different activation functions. SoftMax was the output function. Using images from MNIST and CIFAR-10 databanks, these networks were trained to minimize sparse categorical crossentropy. A large-scale cross-validation was simulated using stochastic Markov chains to account for the law of large numbers for the probability values. Statistical tests support the research hypothesis stating Phish could outperform other activation functions in classification. Future experiments would involve testing Phish in unsupervised learning algorithms and comparing it to more activation functions.

The purpose of this research was to define acceleration in diagnostic procedures for airborne diseases. Airborne pathogenicity can be troublesome to diagnose due to intrinsic variation and overlapping symptoms. Coronavirus testing was an instance of a flawed diagnostic biomarker. The levels of independent variables (IV) were vanilla, sparse, and dense amalgamates formed from multilayer perceptrons and image processing algorithms. The dependent variable (DV) was the classification accuracy. It was hypothesized that if a dense amalgamate is trained to identify Coronavirus, the accuracy would be the highest. The amalgamates were trained to analyze the morphological patches within radiologist-verified medical imaging retrieved from online databanks. Using generative cross-validations, the DV was consulted for each amalgamate. Self-calculated t-tests supported the research hypothesis, with the dense amalgamate achieving 85.37% correct classification rate. The null hypothesis was rejected. Flaws within the databanks were possible sources of error. A new algorithm developed here identified Coronavirus, Mycobacterium, Carcinoma, and Pneumonia from 96-99% accuracy. Future enhancements involve tracking osteopenia/osteoporosis with the algorithm.