With the utilization of deep learning approaches, the key factors for a successful application are sufficient datasets with reliable ground truth, which are generally not easy to obtain, especially in the field of medicine. In recent years, this issue has been commonly addressed with the exploitation of transfer learning via fine-tuning, which enables us to start with a model, pre-trained for a specific task, and then fine-tune (train) only certain layers of the neural network for a related but different target task. However, the selection of fine-tunable layers is one of the major problems of such an approach. Since there is no general rule on how to select layers in order to achieve the highest possible performance, we developed the Differential Evolution based Fine-Tuning ( DEFT ) method for the selection of fine-tunable layers for a target dataset under the given constraints. The method was evaluated against the problem of identifying the osteosarcoma from the medical imaging dataset. The performance was compared against a conventionally trained convolutional neural network, a pre-trained model, and the model trained using a fine-tuning approach with manually handpicked fine-tunable layers. In terms of classification accuracy, our proposed method outperformed the compared methods by a margin of 4.45% to 32.75%.

Phishing stands for a fraudulent process, where an attacker tries to obtain sensitive information from the victim. Usually, these kinds of attacks are done via emails, text messages, or websites. Phishing websites, which are nowadays in a considerable rise, have the same look as legitimate sites. However, their backend is designed to collect sensitive information that is inputted by the victim. Discovering and detecting phishing websites has recently also gained the machine learning community’s attention, which has built the models and performed classifications of phishing websites. This paper presents two dataset variations that consist of 58,645 and 88,647 websites labeled as legitimate or phishing and allow the researchers to train their classification models, build phishing detection systems, and mining association rules.

Nature-inspired algorithms are a very popular tool for solving optimization problems (Yang 2014), (Hassanien and Emary 2016). Numerous variants of nature-inspired algorithms have been developed (Iztok Fister Jr. and Fister 2013) since the beginning of their era. To prove their versatility, those were tested in various domains on various applications, especially when they are hybridized, modified or adapted. However, implementation of nature-inspired algorithms is sometimes a difficult, complex and tedious task. In order to break this wall, NiaPy is intended for simple and quick use, without spending time for implementing algorithms from scratch.

Automatic Classification of Motor Impairment Neural Disorders from EEG Signals Using Deep Convolutional Neural Networks