shivam-5 / pattern-recognition

pattern-recognition

1. MNIST/EMNIST classification: Digit classifier based on CNN architecuture using cross-validation and building final model as ensemble of all best checkpoints. Final accuracy scores of 0.9980 and 0.9981 on MNIST and EMNIST test sets repectively.

2. Activations: Collection of all relevant and useful Activation functions along with their derivatives

3. CNN-layers: Implementation of some CNN layers which include pooling, convolution, batch-normalization. Also contains api to calculate output layer output size

4. DeltaLearning: Implementation of Delta learning algorithm both sequential and batch

5. k-nearest-neighbor: Simple implementation of KNN

6. Linear-disciminant-function: Implemenation of Linear disciminant function alogrithm

7. MLP-1: Implemenation of forward-propagation, backpropagation, weight-update APIs over 1 epoch of input data in Multi layer perceptron

8. NegativeFeedbackLearning: Implementation of basic as wells as Regulatory feedback(also called Decisive Input Modulation) learning

9. Widrow-hoff-sequential: Implementation of widrow-hoff algorithm with sequential learning