Neural Networks Demystified: Part 6
summary
Builds and trains a neural network using the BFGS optimizer via scipy.optimize.minimize, wrapping the network so the objective returns cost and gradients and setting Jacobian=true with a callback to record progress. Training finds a solution with <100 function evaluations (versus ~10^27 by brute force), gives near-zero gradients, and produces a model that predicts test score from hours slept and hours studied where sleep outweighs study — then warns that the model likely overfits.
AI-written summary · about
related talks
01h 25mBuilding & Optimization of Neural Networks from Scratch with Numpy
25mIntroduction to Neural Networks
01h 10mDemystifying Deep Learning for Data Scientists
44mMachine Learning: The Bare Math Behind Libraries
34mBayesian Optimization and it's application to Neural Networks
01h 04mPredicting and optimizing the behavior of large ML models
01h 28mFollow the gradient: an introduction to mathematical optimisation
24mIntelligent system optimizations