Search

Hypertension is becoming a global epidemic for the developing world and continuous blood pressure monitoring and early diagnosis is vital for the prevention of this disease. However, in real life, patients are usually unable to maintain frequent monitoring because of reasons that include forgetfulness, human error and/or machine error. This paper presents a personalized prediction model for blood pressure using Recurrent Kernel Extreme Reservoir Machine (RKERM). This technique combines reservoir computing with RKELM to perform multistep ahead prediction. We use RKERM for blood pressure prediction and its performance is evaluated with other ELM based prediction algorithms. To evaluate our model, we use real world blood pressure data collected from Malaysian population consisting of hypertensive and non-hypertensive patients. The experimental results show that the proposed prediction mechanism has higher prediction accuracy than existing ELM methods.

The major problem on the road to artificial intelligence is the development of lifelong learning systems. They have the ability to learn new concepts incrementally overtime. They are also able to allocate required resources dynamically without human intervention and are able to store data securely. In this work we have extended the incremental classifier and representation learning method known as iCaRL to meet this criterion. The proposed solution is able to learn strong classifiers and a data representation simultaneously. It is able to allocate an optimal scaling plan to meet its resource requirements without human intervention. It securely stores propriety image data by using state of the art interplanetary file system and block chain technology. Finally, it is able to focus on object of interests in an image using attention network. We have shown by experiments on CIFAR-100 and Image net 2012 that it performs better in terms of accuracy than the existing iCaRL system while satisfying criteria of lifelong learning.