A Tour of Deep Learning With C++
summary
Overview of concrete results and tools: demonstrates GANs and convolution math, implements a convolution with cuDNN and MKL, and shows a working edge‑detection run; compiles a TensorFlow custom kernel in ~50 lines and loads it to produce correct sigmoid outputs; trains an image classifier in ~250 C++ lines with MXNet and demos live digit recognition that works; cites scaling numbers: AlexNet ≈ 130 million parameters, ImageNet training 43 days on CPU vs 6 days on a Titan X, and TPU TOPS/TFLOPS comparisons.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:00Task
- 2:17GANZ
- 5:21Layer
- 9:34Computation Graph
- 11:32Parallelization
- 13:51Static vs Dynamic Graphs
- 16:40Machine Learning Operations
- 19:26Kernels
- 22:23Quantization
- 25:22Hardware
- 27:11GPU Example
- 29:21ASIC Example
- 37:46Tensorflow Example
- 42:16Tensorflow
- 44:31PITorch
- 46:28MXN
- 47:23C API
- 52:15Questions
related talks
57mDeep Learning with C++
35mDeconstructing Deep Learning
54mIntroduction to Understanding Deep Learning
01h 04mWhat you always wanted to know about Deep Learning, but were afraid to ask
30mMachine Learning & Neural Networks Demystified
27mDemystifying Deep Neural Networks
40mMachines are Learning
26mDeep Learning in OpenCV