Sleep is a phenomenon that impacts substantial hours of people in 24 hours. Cortical activity is distinct during different phases of sleep. In fact, the gold standard for sleep measurement is polysomnogram (PSG) consisting of electroencephalography (EEG), electromyography (EMG), and electrooculography (EOG). The combination of brainwave activity, muscular activity, and eye movement delineate different stages of sleep in humans. Sleep in small animal models is also possible with EEG/EMG. Classically, sleep is scored by visual examination on the characteristics of the EEG (frequency and amplitude). In mice, sleep is grossly categorized as rapid eye movement (REM) or non-REM (NREM) sleep. REM sleep has unique characteristics for EEG (high frequency, low amplitude, desynchronized waveforms) and EMG (near absent due to muscle atonia). For NREM sleep, the patterns for EEG are highly heterogeneous, making the fine mapping of NREM sleep technically challenging and error-prone among different scorers. NREM sleep is not uniform. The ability to objectively evaluate the subtle stages of NREM sleep allows us to appreciate the function of NREM sleep. My thesis focuses on developing a technique that better characterizes distinct clusters of NREM sleep based on the differential potential of EEG at different frequencies in both normal and disrupted sleep conditions. We proposed that, by mapping NREM sleep to a finer resolution, we will reveal distinct temporal and functional patterns of NREM sleep. We used the machine learning technique, K-mean Clustering, to subcategorized NREM into eight different clusters with distinct EEG patterns. These clusters exhibit temporal specificity at different phases of sleep. Finally, we demonstrated that different NREM clusters are differentially affected by the experimental disruption of sleep. Our method elicits the complexity of NREM sleep, enabling us to investigate whether specific sequences of NREM clusters exist in normal and pathological sleep.