Custom Deep Learning Module programmed in base Python as a learning exercise.
This project showcases a custom DeepLearning module programmed in base Python. This was part of a a Machine Learning course about Deep Learning. The Jupyter Notebook showcases the following:
- Activation functions
- Derived Activation functions
- Loss functions
- Single-Layer Perceptron Classification
- Binominal Logistic Regression Classification
- Multi-Layer Perceptron Classification
- Linear Regression
- Regression using a full Neural Network
Using the module to classify hand-written number from the MNIST dataset
- Classification using a dummy network [step 1-7] with an accuracy of 54%
- Classification using an actual network [step 1-7] with 8.2 minutes training time and 89.9% accuracy.