How layered differentiable programs learn features, dynamics, and generative distributions from data.
Each parameter receives a local derivative of the loss and is updated by gradient descent.
Balances faithful reconstruction with a latent distribution close to a prior.
A generator and discriminator play a minimax game until generated and real samples become difficult to distinguish.
A reusable technical reading framework for moving from vocabulary to working understanding. Use it alongside the formulas and papers below.
How layered differentiable programs learn features, dynamics, and generative distributions from data. Start by naming the representation, objective, and source of evidence in any system you study. This habit separates a useful model explanation from a list of buzzwords.
Use the learning outcomes as checks for understanding: Trace forward and backward passes through common architectures; Understand convolution, recurrence, residual paths, and latent variables; Diagnose vanishing gradients, overfitting, and mode collapse. For each claim, ask what assumptions make it true, what data it needs, and how it could fail.
Translate each equation into a small experiment before treating it as memorized knowledge. Backpropagation update: Each parameter receives a local derivative of the loss and is updated by gradient descent. VAE evidence lower bound: Balances faithful reconstruction with a latent distribution close to a prior. GAN objective: A generator and discriminator play a minimax game until generated and real samples become difficult to distinguish.
Read primary work with a repeatable lens: problem, method, data, measurement, limits, and what would change your mind. This section starts with “Deep Residual Learning for Image Recognition”, “Generative Adversarial Nets”, “Auto-Encoding Variational Bayes”. Follow citations outward only after you can explain the central claim in your own words.
Implement a minimal example using Backpropagation update; record inputs, outputs, and a failure case.
Define a baseline, a success metric, a stress test, and the evidence required before you trust an improvement.
Choose one paper from the library, reproduce a small claim, then write down the questions its evaluation leaves open.
Primary sources and peer-reviewed work to deepen this page.
K. He · X. Zhang · S. Ren · J. Sun
Introduces residual skip connections, enabling much deeper networks to optimize reliably.
I. Goodfellow et al.
Frames generation as a two-player minimax game between a generator and discriminator.
D. P. Kingma · M. Welling
Makes variational inference trainable with stochastic gradient methods and the reparameterization trick.