The tropical cyclone formation process is one of the most complex natural phenomena which is governed by various atmospheric, oceanographic, and geographic factors that varies with time and space. Despite several years of research, accurately predicting tropical cyclone formation remains a challenging task. While the existing numerical models have inherent limitations, the machine learning models fail to capture the spatial and temporal dimensions of the causal factors behind TC formation. In this study, a deep learning model has been proposed that can forecast the formation of a tropical cyclone with a lead time of up to 60 hours with high accuracy. The model uses the high-resolution reanalysis data ERA5 (ECMWF reanalysis 5th generation), and best track data IBTrACS (International Best Track Archive for Climate Stewardship) to forecast tropical cyclone formation in six ocean basins of the world. For 60 hours lead time the models achieve an accuracy in the range of 86.9% - 92.9% across the six ocean basins. The model takes about 5-15 minutes of training time depending on the ocean basin, and the amount of data used and can predict within seconds, thereby making it suitable for real-life usage.

Deep learning researchers have a keen interest in proposing two new novel activation functions which can boost network performance. A good choice of activation function can have significant consequences in improving network performance. A handcrafted activation is the most common choice in neural network models. ReLU is the most common choice in the deep learning community due to its simplicity though ReLU has some serious drawbacks. In this paper, we have proposed a new novel activation function based on approximation of known activation functions like Leaky ReLU, and we call this function Smooth Maximum Unit (SMU). Replacing ReLU by SMU, we have got 6.22% improvement in the CIFAR100 dataset with the ShuffleNet V2 model.

Well-known activation functions like ReLU or Leaky ReLU are non-differentiable at the origin. Over the years, many smooth approximations of ReLU have been proposed using various smoothing techniques. We propose new smooth approximations of a non-differentiable activation function by convolving it with approximate identities. In particular, we present smooth approximations of Leaky ReLU and show that they outperform several well-known activation functions in various datasets and models. We call this function Smooth Activation Unit (SAU). Replacing ReLU by SAU, we get 5.12% improvement with ShuffleNet V2 (2.0x) model on CIFAR100 dataset.

An activation function is a crucial component of a neural network that introduces non-linearity in the network. The state-of-the-art performance of a neural network depends on the perfect choice of an activation function. We propose two novel non-monotonic smooth trainable activation functions, called ErfAct and PSerf. Experiments suggest that the proposed functions improve the network performance significantly compared to the widely used activations like ReLU, Swish, and Mish. Replacing ReLU by ErfAct and PSerf, we have 5.21% and 5.04% improvement for top-1 accuracy on PreactResNet-34 network in CIFAR100 dataset, 2.58% and 2.76% improvement for top-1 accuracy on PreactResNet-34 network in CIFAR10 dataset, 1.0%, and 1.0% improvement on mean average precision (mAP) on SSD300 model in Pascal VOC dataset.

The weather-related forecast is one of the difficult problems to solve due to the complex interplay between various cause factors. Accurate tropical cyclone intensity prediction is one such problem that has huge importance due to its vast social and economic impact. Cyclones are one of the devastating natural phenomena that frequently occur in tropical regions. Being a tropical region, Indian coastal regions are frequently affected by tropical cyclones [1] that originate into the Arabian Sea (AS) and Bay of Bengal (BOB), which are parts of the North Indian Ocean (NIO). With the increasing frequency of cyclones in NIO [2], it becomes more crucial to develop a model that can forecast the intensity of a cyclone for a longer period of time by observing the cyclone only for a small period of time. Various statistical and numerical methods have been developed to predict the intensity of cyclones [3-7] but all these methods lack effectiveness in terms of accuracy and computation time.

Deep networks are constructed with multiple hidden layers and neurons. Non-linearity is introduced in the network via activation function in each neuron. ReLU [1] is proposed by Nair and Hinton and is the favourite activation in the deep learning community due to its simplicity. Though ReLU has a drawback called dying ReLU, and in this case, up to 50% neurons can be dead due to vanishing gradient problem, i.e. there are numerous neurons which has no impact on the network performance. To overcome this problem, later Leaky Relu [2], Parametric ReLU [3], ELU [4], Softplus [5] was proposed, and they have improved the network performance though it's still an open problem for researchers to find the best activation function. Recently Swish [6] was found by a group of researchers from Google brain, and they used automated searching technique. Swish has shown some improvement in accuracy over ReLU.

TC is characterised by warm core, and a low and availability of huge data, new models using Artificial pressure system with a large vortex in the atmosphere. TC Neural Networks (ANNs) have been increasingly used to brings strong winds, heavy precipitation and high tides in forecast track and intensity of cyclones (Leroux et al. 2018; coastal areas and resulted in huge economic and human loss. Alemany et al. 2018; Giffard-Roisin et al. 2020; Moradi Kordmahalleh, Over the years, many destructive TCs have originated in the Gorji Sefidmazgi, and Homaifar 2016). North Indian Ocean (NIO), consisting of the Bay of Bengal The most important prediction about a TC is its arrival at and the Arabian Sea. In 2008, Nargis, one of the disastrous land, known as landfall of a cyclone. The accurate prediction TC in recent times, originated in the Bay of Bengal and resulted about the location and time of the landfall, and intensity of in 13,800 casualties alone in Myanmar and caused the cyclone at the landfall will hugely help authorities to take US$15.4 billion economic loss (Fritz et al. 2009). In 2018, preventive measures and reduce material and human loss. In Fani cyclone caused 89 causalities in India and Bangladesh, this work, we attempt to predict intensity, location, and time and US$9.1 billion economic loss (Kumar, Lal, and Kumar of the landfall of a TC at any instance of time during the 2020).

Landfall of a tropical cyclone is the event when it moves over the land after crossing the coast of the ocean. It is important to know the characteristics of the landfall in terms of location and time, well advance in time to take preventive measures timely. In this article, we develop a deep learning model based on the combination of a Convolutional Neural network and a Long Short-Term memory network to predict the landfall's location and time of a tropical cyclone in six ocean basins of the world with high accuracy. We have used high-resolution spacial reanalysis data, ERA5, maintained by European Center for Medium-Range Weather Forecasting (ECMWF). The model takes any 9 hours, 15 hours, or 21 hours of data, during the progress of a tropical cyclone and predicts its landfall's location in terms of latitude and longitude and time in hours. For 21 hours of data, we achieve mean absolute error for landfall's location prediction in the range of 66.18 - 158.92 kilometers and for landfall's time prediction in the range of 4.71 - 8.20 hours across all six ocean basins. The model can be trained in just 30 to 45 minutes (based on ocean basin) and can predict the landfall's location and time in a few seconds, which makes it suitable for real time prediction.

Artificial neural networks (ANNs) have occupied the center stage in the realm of deep learning in the recent past. ANNs are made up of several hidden layers, and each hidden layer consists of several neurons. At each neuron, an affine linear map is composed with a nonlinear function known as activation function. During the training of an ANN, the linear map is optimized, however an activation function is usually fixed in the beginning along with the architecture of the ANN. There has been an increasing interest in developing a methodical understanding of activation functions, in particular with regards to the construction of novel activation functions and identifying mathematical properties leading to a better learning [1]. An activation function is considered good if it can increase the learning rate and leaning to better convergence which leads to more accurate results. At the early stage of deep learning research, researchers used shallow networks (fewer hidden layers), and tanh or sigmoid, were used as activation functions. As the research progressed and deeper networks (multiple hidden layers) came into fashion to achieve challenging tasks, Rectified Linear Unit (ReLU)([2], [3], [4]) emerged as the most popular activation function. Despite its simplicity, deep neural networks with ReLU have learned many complex and highly nonlinear functions with high accuracy.