Oksvabram / Cognitive-Computing-final-project

Exploration of various ML models and techniques for cognitive computing tasks. The primary focus is analysing hidden representations and the effectiveness in classifying data

Cognitive Computing final project

Exploration of various ML models and techniques for cognitive computing tasks. The primary focus is analysing hidden representations and the effectiveness in classifying data

Models are tested on the Fashion MNIST dataset

Key components:

• Deep Belief Network: implementation and analysis of DBNs, extraction and clustering of hidden representations
• Dendrogram Analysis: visualization of hierarchical clustering of hidden representations from DBN
• Linear Read-out: decoding hidden representations using linear classifiers to assess the information contained in each hidden layer
• Convolutional Neural Networks: application of CNNs to compare their performance with DBN on the same dataset

Highlights

• Data analysis: detailed exploration of hidden layer representations and their clustering using dendrograms
• Model training: training and evaluation of linear models and CNNs to classify image data
• Comparative Analysis: comparison of the performance of DBNs and CNNs