SASDL and RBATQ: Sparse Autoencoder With Swarm Based Deep Learning and Reinforcement Based Q-Learning for EEG...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Electroencephalography (EEG) is crucial for analyzing brain electrical activity and diagnosing neurological disorders.
• Deep learning offers automated feature engineering, surpassing traditional machine learning in complexity.
• Swarm intelligence techniques are effective for solving complex, non-linear problems.

Purpose of the Study:

• To propose two advanced deep learning methods for accurate classification of epilepsy and schizophrenia from EEG data.
• To leverage the strengths of deep learning and swarm intelligence for improved diagnostic capabilities.

Main Methods:

• Developed Sparse Autoencoder with swarm-based deep learning (SASDL) using Particle Swarm Optimization (PSO), Cuckoo Search Optimization (CSO), and Bat Algorithm (BA).
• Introduced a Reinforcement Learning-based method (RBATQ) integrating Bidirectional Long-Short Term Memory (BiLSTM), Attention Mechanism, Tree LSTM, and Q-learning.

Main Results:

• Both novel deep learning techniques demonstrated high performance on epilepsy and schizophrenia EEG datasets.
• Classification accuracy exceeding 93% was achieved across all tested datasets for both methods.

Conclusions:

• The proposed SASDL and RBATQ techniques show significant potential for accurate EEG-based classification of neurological disorders.
• These advanced deep learning approaches offer a robust framework for diagnosing conditions like epilepsy and schizophrenia.