Deep Learning for Developers
summary
Concrete claim: a character‑level convolutional model detects programming language from code files using Keras on top of TensorFlow. Architecture uses stacked convolutions (kernels 3,5,9,19), pooling, dropout, inception‑style concatenation and softmax over 16 programming languages. Training uses about 2k examples per language with 1k test and runs on GPU (≈8 hours). The model classifies full files reliably in demos but shows limited generalization on short snippets and needs substantially more data to reach robust accuracy.
AI-written summary · about
chapters (41)
- 0:00Introduction
- 0:30Ali Kheyrollahi
- 1:06Road Marking and Text Recognition
- 2:37What is Deep Learning
- 3:19Agenda
- 4:02Realistic Example
- 5:02History of Deep Learning
- 5:45Story Time
- 8:41Living together
- 8:57McCulloch
- 9:45Wiener
- 10:18Weiner
- 11:05Trillo
- 11:37Rosenblatt
- 12:50Recap
- 13:36Perceptron
- 14:00Sum
- 14:55Activation Functions
- 15:17Marvin and Frank Rosenblatt
- 16:10The First Winter
- 16:43Geoff Hinton
- 17:21Yan Laocoon
- 17:51Convolutional Neural Network
- 19:34Feature extraction
- 20:55Overfitting
- 21:58Support Vector Machine
- 23:03Neural Networks
- 24:01Yan Lac Kun
- 24:34Ben Jia
- 24:57Gradient Descent
- 28:18ImageNet
- 29:06ZFNet
- 29:28VGGNet
- 30:35Inception Module
- 31:41ResNet
- 32:20BackPropagation
- 32:38Learning is hard
- 32:55Programming Language Detection
- 36:51Architecture
- 38:40Max pooling
- 39:16Demo
related talks
35mDeconstructing Deep Learning
47mDeep Learning: a practical introduction from an Android dev perspective
57mDeep Learning with C++
55mA Tour of Deep Learning With C++
01h 06mDeep Learning with TensorFlow
37mCode First, Math Later: Learning Neural Nets Through Implementation and Examples
29mAutomating Software Development with Deep Learning
38mDeep Learning for detection on a phone