Deep learning Framework

From input to output, all steps and methods are presented for generic classification.

1. Data
2. Preprocessing
   1. Normalization
   2. Augmentation
3. Model
   1. Parameter Initialization
   2. Regularization (L1, L2, Dropout, Batch Normalization, Early Stopping, Layer Normalization)

4. Loss

5. Training
   1. Backpropagation
   2. Gradient Descent (Vanila, Mini-batch, Stochastic)
   3. Optimizer (SGD, Momentum, Nesterov Momentum, Adagrad, RMSProp, Adam, AdaDelta, AdaMax, Nadam, AMSGrad)
6. Evaluation
   1. Metrics (Accuracy, Precision, Recall, F1, AUC, ROC, PR, Log Loss, Confusion Matrix, IoU, Dice, GIoU, mAP, mIoU, mDice, mGIoU, mRecall, mPrecision, mF1, mAccuracy, mLog Loss, mConfusion Matrix)
   2. Visualization
   3. Latency and Scalability (Inference Time, Throughput, Latency)