Deep neural networks suffer from a phenomenon called internal covariate shift, where the distribution of inputs to each layer changes continuously during training as the weights of preceding layers are updated. This instability forces lower learning rates, creates vanishing gradient problems, and dramatically increases training time.
Without normalization, networks with many layers struggle because gradients become exponentially smaller as they propagate backward, neurons in intermediate layers receive inputs with increasingly wild distributions, and the optimization landscape becomes highly non-convex and difficult to navigate.
Batch Normalization (BN), introduced by Ioffe and Szegedy in 2015, addresses this by normalizing layer inputs to have zero mean and unit variance within mini-batches, allowing much higher learning rates and faster convergence.
Layer Normalization (LN), proposed by Ba et al., offers an alternative that normalizes across features rather than batch samples, making it independent of batch size and particularly valuable for recurrent networks and transformer architectures. Understanding both techniques is critical because they fundamentally change how information flows through networks and enable training of deeper, more complex models that were previously intractable.
Analogy🏏Cricket
🏏 Think of it like cricket: Imagine Virat Kohli batting in a Test match innings—each delivery he faces builds on the context of all previous deliveries in that innings. The bowler's strategy evolves based on what happened in earlier overs; Kohli's mental state and approach shift based on the match situation, the bowler's previous deliveries, and the scoring rate. His decision to play an aggressive shot or defend depends entirely on this accumulated context—information from the past 50 deliveries that his mind actively maintains. Now map this to an RNN: each timestep is like one delivery Kohli faces, the input is the ball characteristics, the hidden state is Kohli's accumulated mental model of the bowler and match situation, and the output is his batting decision for that delivery. The recurrent connection is Kohli carrying forward his understanding from delivery 1 through delivery 2, 3, 4... all the way to delivery 50—he never resets this knowledge. However, vanilla RNNs suffer a critical problem: like a batsman whose memory of early overs fades by the 50th over (vanishing gradient), the network forgets distant context. LSTMs fix this like Kohli maintaining a written scorecard—explicit gates (input gate, forget gate, output gate) are like decision checkpoints where he consciously updates what he remembers (forget gate), what new information to integrate (input gate), and what to use for his next shot (output gate). This gating mechanism prevents information decay, allowing Kohli to maintain crucial context from delivery 1 even when deciding his shot on delivery 50. Understanding RNNs and LSTMs reveals why sequential problems fundamentally require mechanisms to preserve and selectively use historical information—just as Kohli's effectiveness depends on never losing track of the match narrative.
🏏 Showing the Cricket analogy — a Cricket version isn’t available for this concept yet.