Search

In this work we study the behavior of Prefrontal Cortex (PFC) and understand its role in task switching by developing a biologically based computational model. We build the PFC neurons using Spiking Neural Networks (SNN) with biologically realizable features having lateral inhibition, synaptic weight changes using unsupervised Spike Timing Dependant Plasticity (STDP) learning rule, spiking threshold and biological ranges for neuronal parameter values. The SNN is composed of Leaky Integrate and Fire (LIF) neurons which are efficient to model and represents the Excitatory neurons in Glutamate layer and Inhibitory neurons in GABA layer. In this implementation we use two real world datasets as tasks for the PFC network to learn. We demonstrate the switching behavior of the neurons and their synaptic weight adaptations by formulating experiments in a manner consistent with real world trials used in the study of cognitive psychology. Using these experiments we show how our model adapts and responds to task changes exhibiting biological behaviors like Long Term Potentiation (LTP), Long Term Depression (LTD) and Task-set reconfiguration (TSR) thereby giving insights into understanding the importance of duration between changing tasks and its effect on performance and efficacies of multi-tasking. The results shown in this paper relate favorably well with the natural neuronal responses found in the brain.

Data analysis is a demanding task that involves extracting deep insights hidden in data. Many businesses enforce data analysis irrespective of the domain, as it is crucial in minimizing developmental risks. Raw data cannot be used to perform any analysis as poor-quality data leads to erroneous decision-making. This makes data quality assessment a necessary function before data analysis. Data quality is a multi-dimensional factor that affects the analysis in multiple ways. Among all the dimensions, consistency is one of the most critical dimensions to assess. Context of data plays an important role in consistency assessment, as the records are inherently related within a dataset. Existing studies are computationally expensive and do not consider the context of data. In this paper, we propose a comprehensive context-aware data consistency assessment tool that uses machine learning to evaluate the consistency of data. Our model was developed on Apache Hadoop and Apache Spark to support big data, as well as to boost some computationally intensive algorithms.

Data analysis is the most important aspect of any business as it is critical to decision-making. Data quality assessment is a necessary function to be performed before data analysis, as the quality of data has high impact on the outcome of the analysis. Data quality is a multi-dimensional factor that affects the analysis in numerous ways. Among all the dimensions, accuracy is the most important and hardest dimension to assess. With the advent of big data, this problem becomes more complicated. There are only few studies that focus on data accuracy with minimal domain expert dependency. In this paper, we propose an extensive data accuracy assessment tool that uses machine learning to determine the accuracy of data. In addition, our model also addresses the intrinsic and contextual categories of data accuracy. Our model was developed on Apache Spark which serves as the big data environment for handling large datasets.

The concept of modeling a machine which can adapt to the dynamic changes in environment has fascinated the field of Artificial Intelligence. Machine Learning has made inroads in every possible domain. New techniques are developed which can mimic human like responses and thoughts. Cognitive computing has developed renewed interest in the community with advent of Artificial Neural Nets (ANN). In this paper, we present a biological inspired approach to building a augmented memory recall model which can learn usage access patterns and reconstruct from them when presented with noisy or broken concepts. We use Hopfield Networks in a distributed parallel architecture like Hadoop. We also present a mechanism for augmenting the memory capacity of Hopfield Nets. Our model is tested on a real world dataset by parallelizing the learning process thereby increasing the computing power to recognize patterns.