More Resources
Scala Models
BigDL provides loads of popular models ready for use in your application. Some of them are listed blow. See all in scala neural network models.
- LeNet: it demonstrates how to use BigDL to train and evaluate the LeNet-5 network on MNIST data.
- Inception: it demonstrates how to use BigDL to train and evaluate Inception v1 and Inception v2 architecture on the ImageNet data.
- VGG: it demonstrates how to use BigDL to train and evaluate a VGG-like network on CIFAR-10 data.
- ResNet: it demonstrates how to use BigDL to train and evaluate the ResNet architecture on CIFAR-10 data.
- RNN: it demonstrates how to use BigDL to build and train a simple recurrent neural network (RNN) for language model.
- Auto-encoder: it demonstrates how to use BigDL to build and train a basic fully-connected autoencoder using MNIST data.
Scala Example
BigDL ships plenty of Scala examples to show how to use BigDL to solve real problems. Some are listed blow. See all of them in scala deep learning examples
- Text Classification: it demonstrates how to use BigDL to build a text classifier using a simple convolutional neural network (CNN) model.
- Image Classification: it demonstrates how to load a BigDL or Torch model trained on ImageNet data (e.g., Inception or ResNet), and then applies the loaded model to classify the contents of a set of images in Spark ML pipeline.
- Load Model: it demonstrates how to use BigDL to load a pre-trained Torch or Caffe model into Spark program for prediction.
- ML Pipline: it demonstrates how to use BigDL DLClassifier to train a Logistic Regression Model. DLClassifier extends Spark Estimator and can act as a stage in a ML Pipeline.
- TreeLSTM Sentiment: it demonstrates how to use BigDL train a model on Stanford Treebank dataset dataset using binary TreeLSTM and Glove word embedding vectors.
- UDF Predictor: it demonstrates how to load BigDL model as UDF to perform predictions in Spark SQL/Dataframes.