TensorFlow in Three Sentences
summary
Three clear claims: build a computational graph (constants, placeholders, variables; session run), choose a level of abstraction from raw ops to DNNClassifier with 1024 hidden units or tf.layers, and join the ecosystem. Demos include linear fit, 784→10 softmax with batch size 100 for 1,000 steps, a convolutional model and an Inception v3 served in Docker, yielding ~97% on MNIST variants while the Inception demo mislabels a dog as a red ball. Tools: TensorFlow, tf.layers/estimators, TensorBoard, Docker/Kata Coda.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 4:58Natural Neural Networks
- 9:04Build a Computational Graph
- 12:06Running a Computational Graph
- 15:03Placeholders
- 19:02Loss Function
- 21:00Classification Tasks
- 23:35Computational Graph
- 31:35Different levels of abstraction
- 33:58Deep neural network classifier
- 37:44Convolutional neural networks
- 40:12Building a convolutional network
- 47:21Recap
- 52:34ImageNet
related talks
44mTensorflow, Tensorflow, What is it that you see?
41mLearning TensorFlow
55mTensorflow: Learning Machines
45mThe most soft and fluffy intro into Machine Learning and Deep Neural Networks
31mTensorflow for Janitors
37mSimplifying Training Deep & Serving Learning Models...using Tensorflow
57mDeep Learning with C++
38mThe Hitchhiker's Guide to Tensorflow: from Tensors to running models in production