The MOST in-depth look at neural network theory for machine learning, with both pure Python and Tensorflow code
InstructorLazy Programmer Inc.
Coupon
The curriculum builds entire feedforward neural networks from scratch using Python and NumPy, covering the mathematical implementation of backpropagation.
Instruction includes specific components such as categorical cross-entropy loss, derivatives of nonlinearities, and gradient descent tutorials.
Editorial course preview
These 4 complementary views highlight concrete, legible examples from the course presentation.
This slide outlines the course objectives, promising to demystify deep neural networks through derivation and implementation using only basic Python, accompanied by a Richard Feynman quote.
This slide contrasts superficial breadth with in-depth fundamentals by illustrating how most courses focus on memorizing code while this one digs deeper into neural networks.
This slide promotes a machine learning course that prioritizes building from basic first principles using real programming and math rather than relying on libraries like SciKit-Learn.
This slide contrasts basic three-line deep learning APIs with the course's focus on building state-of-the-art models, displaying sample Keras code alongside explanatory text.
Practical application is provided through an e-commerce data project and a facial expression recognition task.
The depth of calculus required may exceed the stated prerequisites, according to a signal from a review created before the displayed update date; note that the update label does not prove a correction has been made.
Some mathematical equations lack explicit explanations of variables or Greek symbols, according to a signal from a review created before the displayed update date; note that the update label does not prove a correction has been made.
The curriculum contains mentions of Python 2 vs Python 3 and specific discussions regarding the year 2020, suggesting potential technical aging.
The mismatch between stated prerequisites and the actual calculus depth required impacts suitability for some learners. This signal predates the displayed course update; the update may have addressed it, but the update label does not prove that it was corrected.