The term Artificial Neural Networks (ANN) refers to a set of algorithms that try to mimic the brain. It is constructed of an input layer, the data you input, a number of hidden layers, which help to group unlabelled data according to similarities, and then the output layer which is the hypothesis.
Neural networks can fall into a range of categories, this blog will focus on classification.
Classification problems involve either binary decisions or multiple-class identification in which observations are separated into categories according to specified characteristics. For example, whether a cancer is benign or malignant, whether an object is a dog, or classing a letter as a particular character in the alphabet. The ANN is expected to build a model that can solve and learn from a particular dataset and for this model to be implemented in similar cases.
The NHS recently launched their latest campaign 'Cervical Screening Saves Lives' which urges women to attend their cervical screening, with data showing that the number of women attending screening has fallen to a 20 year low. Statistics show that around 2,600 women are diagnosed with cervical cancer in England each year and around 690 women die from the disease. It is estimated that if everyone attended screening regularly, 83% of cervical cases could be prevented.
This very high prevention rate comes down to the role artificial neural networks (ANN) plays in many medical imaging applications. The detection of cervical cancer cells uses an ANN for classifying the normal and abnormal cells in the cervix region of the uterus. Cervical cancer detection is very challenging because this cancer occurs without any symptoms. The classification between the normal, abnormal and cancerous cells is identified by using an artificial neural network which produces more accurate results than the manual screening methods like Pap smear test. The ANN uses several architectures for easy and accurate detection of cervical cells.
Handwriting recognition is one of the most energetic and challenging research areas in image processing and pattern recognition. In the instance of reading government documents, an effort can be made to recognise handwritten characters for english alphabets with feature extraction using Multilayer Feed Forward Neural Network. Recognising the text of a document would be useful in many diverse applications like post code, bank cheques and other official documents. It also finds uses in detective or police departments in applications like handwriting based person identification, identifying real from forged documents, etc and then stores them. It contributes immensely to the advancement of automation process and improves the interface between man and machine in numerous applications.
ANNs can be used to protect organisations from several types of attacks, such as Denial-of-service attack (DDoS) and malicious software. Malware itself is a huge problem, with at least 325,000 new malicious files being generated every day. Yet, no more than 10 percent of the files change from iteration to iteration, so algorithm-based learning models that can predict these variations are able to detect which files are malware with amazing accuracy. Neural networks could be used to detect any change or anomaly in network traffic to identify potentially malicious activities such as brute force attacks, unusual failed logins and file extraction.
The Anaeko team are committed to finding the best model suited to your data problem. To learn more about how we can help your company or department, contact us below;