TensorFlow implementation of Value Iteration Networks (VIN): Clean, Simple and Modular

7 months after

VIN: Value Iteration Networks

This is an implementation of Value Iteration Networks (VIN) in TensorFlow to reproduce the results.(PyTorch version)

Architecture of Value Iteration Network

Key idea

  • A fully differentiable neural network with a 'planning' sub-module.
  • Value Iteration = Conv Layer + Channel-wise Max Pooling
  • Generalize better than reactive policies for new, unseen tasks.

Learned Reward Image and Its Value Images for each VI Iteration

Visualization Grid world Reward Image Value Images


This repository requires following packages:

  • Python >= 3.6
  • Numpy >= 1.12.1
  • TensorFlow >= 1.0
  • SciPy >= 0.19.0


Each data sample consists of (x, y) coordinates of current state in grid world, followed by an obstacle image and a goal image.

Dataset size 8x8 16x16 28x28
Train set 77760 776440 4510695
Test set 12960 129440 751905

Running Experiment: Training

Grid world 8x8

python --datafile data/gridworld_8x8.npz --imsize 8 --lr 0.005 --epochs 30 --k 10 --batch_size 128

Grid world 16x16

python --datafile data/gridworld_16x16.npz --imsize 16 --lr 0.008 --epochs 30 --k 20 --batch_size 128

Grid world 28x28

python --datafile data/gridworld_28x28.npz --imsize 28 --lr 0.003 --epochs 30 --k 36 --batch_size 128


  • datafile: The path to the data files.
  • imsize: The size of input images. From: [8, 16, 28]
  • lr: Learning rate with RMSProp optimizer. Recommended: [0.01, 0.005, 0.002, 0.001]
  • epochs: Number of epochs to train. Default: 30
  • k: Number of Value Iterations. Recommended: [10 for 8x8, 20 for 16x16, 36 for 28x28]
  • ch_i: Number of channels in input layer. Default: 2, i.e. obstacles image and goal image.
  • ch_h: Number of channels in first convolutional layer. Default: 150, described in paper.
  • ch_q: Number of channels in q layer (~actions) in VI-module. Default: 10, described in paper.
  • batch_size: Batch size. Default: 128



Performance: Test Accuracy

NOTE: This is the accuracy on test set. It is different from the table in the paper, which indicates the success rate from rollouts of the learned policy in the environment.

Test Accuracy 8x8 16x16 28x28
TensorFlow 99.03% 90.2% 82%
PyTorch 99.16% 92.44% 88.20%

Speed with GPU

Speed per epoch 8x8 16x16 28x28
TensorFlow 4s 25s 165s
PyTorch 3s 15s 100s

Frequently Asked Questions

  • Q: How to get reward image from observation ?

    • A: Observation image has 2 channels. First channel is obstacle image (0: free, 1: obstacle). Second channel is goal image (0: free, 10: goal). For example, in 8x8 grid world, the shape of an input tensor with batch size 128 is [128, 2, 8, 8]. Then it is fed into a convolutional layer with [3, 3] filter and 150 feature maps, followed by another convolutional layer with [3, 3] filter and 1 feature map. The shape of the output tensor is [128, 1, 8, 8]. This is the reward image.
  • Q: What is exactly transition model, and how to obtain value image by VI-module from reward image ?

    • A: Let us assume batch size is 128 under 8x8 grid world. Once we obtain the reward image with shape [128, 1, 8, 8], we do convolutional layer for q layers in VI module. The [3, 3] filter represents the transition probabilities. There is a set of 10 filters, each for generating a feature map in q layers. Each feature map corresponds to an "action". Note that this is larger than real available actions which is only 8. Then we do a channel-wise Max Pooling to obtain the value image with shape [128, 1, 8, 8]. Finally we stack this value image with reward image for a new VI iteration.


Further Readings