← 📊 Evaluation & ML FoundationsNEXT IN EVALUATION & ML FOUNDATIONSCNNs: Convolution, Pooling, Receptive Fields→
Core
Training Neural Nets: Init, Normalization, Dropout, LR Schedules
The practical recipe that makes deep nets train at all: scale-aware weight initialization (Xavier, He), normalization layers (batch, layer, RMS) that keep activations well-conditioned, dropout as stochastic regularization, and warmup plus cosine learning-rate schedules. Applied AI interviews probe this because picking the wrong init or norm is a common reason training diverges or plateaus, and knowing why each helps separates people who have trained models from people who have only called .fit().
a free account unlocks the core curriculum tier · no card
RELATED CONCEPTS
PRACTICE THIS IN REAL QUESTIONS
Machine Learning & Data ScienceWhat is batch normalization, why does it help training, and how does it differ at train vs inference?→Machine Learning & Data ScienceWhat is dropout, and how does it regularize a neural network?→Machine Learning & Data ScienceExplain backpropagation. Walk through the chain rule for a simple two-layer network.→Machine Learning & Data ScienceWhat causes vanishing and exploding gradients, and how do activations, initialization, and residuals fix them?→Coding & DSAImplement a dropout layer (train and inference) from scratch.→Machine Learning & Data ScienceCompare SGD, momentum, RMSProp, Adam, and AdamW. Why does AdamW decouple weight decay?→
COMPANIES THAT ASSUME THIS