LURS 2019

Ariel Feldman

Advised by Dr. Caleb Kemere

A Machine Learning Approach to Predicting Occurrence of Sharp-Wave Ripple Complexes

Figure credit: Etienne Ackermann

What Are Sharp-Wave Ripples?

Colgin et al. Nature Reviews, 2016

Why Do We Need Prediction?

Missing approx. 40 - 60% of the SWR
- Never interrogated ripple generation
- Validation of in vivo interaction
Research context
- Interact with memory recall & consolidation
- Analyze impact on decision-making

Objective

Design Challenges

Goal: Apply machine learning techniques towards building a realtime closed loop SWR disruption system.

Anomalous event
- Ripples happen rarely
- Complex classification problem
Loss value
- Weighting loss misclassifications
- Custom function
Must be realtime applicable

Spectral Domain

Sensitivity

Convolutional vs. Recurrent

\frac{TP}{TP + FN}

\frac{TN}{TN + FP}

Convolutional Architecture

Multi-Layer Perceptron

CA1

CA2

CA3

Weighting Misclassification

def sensitivity(y_true, y_pred):
    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
    possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))
    return true_positives / (possible_positives + K.epsilon())
def specificity(y_true, y_pred):
    true_negatives = K.sum(K.round(K.clip((1-y_true) * (1-y_pred), 0, 1)))
    possible_negatives = K.sum(K.round(K.clip(1-y_true, 0, 1)))
    return true_negatives / (possible_negatives + K.epsilon())

t = 1
class_weight={
    0:  1,
    1:  len(np.where(dfy_train == 0)[0])/ len(np.where(dfy_train == 1)[0]) * t
}

How Will We Approach This?

Ensemble/Cascade Learning
- Multiple mediocre classifiers
Incorporation of Long Short Term Memory
- Neural signals are time-series
  - temporal importance

Ensemble Classifiers

Figures from Analytics Vidhya

Long Short Term Memory Networks

Figure from MDPI and ACS Style

Why Do We Care?

Memory Enhancement
- Cortical p-waves
Greater Understanding of Consolidation
- How do we improve learning?
- Neural basis of memory disorders
Post-Traumatic Stress Disorder
- Remove access to specific memories

Acknowledgments

Dr. Caleb Kemere

Shayok Dutta

PhD Candidate, Collaborator

Principal Investigator