Google to Improve Mobile Machine Learning Models

The team used multi-criteria optimization to achieve high speed and accuracy in the neural network MnasNet
09 August 2018   1039

The neural network MnasNet uses a training method with reinforcement to select the appropriate mobile device architecture.

The MnasNet system contains:

  • object management based on a recurrent neural network for learning and selecting the architecture of the model;
  • he coach who builds the models and teaches their accuracy;
  • an output machine that calculates the speed of the model on the phones using TensorFlow.

The team used multi-criteria optimization to achieve high speed and accuracy. In addition, scientists used the learning algorithm with reinforcement and remuneration function. Thus, MnasNet finds Pareto optimality for each platform. For each mobile device, individual features are created that require a specific architecture.

Automated Neural Design Approach
Automated Neural Design Approach

To achieve the optimal balance between the flexibility of search and the scope of feasible solutions, a hierarchical approach has been applied. It represents a convolutional neural network in the form of a series of blocks. After, sequential search to assign layers of architecture to each block is being used. Due to this, each layer uses different operations and connections. In this case, the layers of blocks are copied by force.

MnasNet blocks
MnasNet blocks

Scientists have included the search speed of architecture in the function of rewarding the search algorithm.

MnasNet Blocks
MnasNet Blocks

According to the developers, the system picks up the model 1.5 times faster than MobileNetV2 and 2.4 times faster than NASNet, while taking into account the optimal accuracy and speed. As a proof, the result was demonstrated by the example of working with ImageNet image database.

Nvidia to Open SPADE Source Code

SPADE machine learning system creates realistic landscapes based on rough human sketches
15 April 2019   674

NVIDIA has released the source code for the SPADE machine learning system (GauGAN), which allows for the synthesis of realistic landscapes based on rough sketches, as well as training models associated with the project. The system was demonstrated in March at the GTC 2019 conference, but the code was published only yesterday. The developments are open under the non-free license CC BY-NC-SA 4.0 (Creative Commons Attribution-NonCommercial-ShareAlike 4.0), allowing use only for non-commercial purposes. The code is written in Python using the PyTorch framework.

Sketches are drawn up in the form of a segmented map that determines the placement of exemplary objects on the scene. The nature of the generated objects is set using color labels. For example, a blue fill turns into sky, blue into water, dark green into trees, light green into grass, light brown into stones, dark brown into mountains, gray into snow, a brown line into a road, and a blue line into the river. Additionally, based on the choice of reference images, the overall style of the composition and the time of day are determined. The proposed tool for creating virtual worlds can be useful to a wide range of specialists, from architects and urban planners to game developers and landscape designers.

Objects are synthesized by a generative-adversarial neural network (GAN), which, based on a schematic segmented map, creates realistic images by borrowing parts from a model previously trained on several million photographs. In contrast to the previously developed systems of image synthesis, the proposed method is based on the use of adaptive spatial transformation followed by transformation based on machine learning. Processing a segmented map instead of semantic markup allows you to achieve an exact match of the result and control the style.

To achieve realism, two competing neural networks are used: the generator and the discriminator (Discriminator). The generator generates images based on mixing elements of real photos, and the discriminator identifies possible deviations from real images. As a result, a feedback is formed, on the basis of which the generator begins to assemble more and more qualitative samples, until the discriminator ceases to distinguish them from the real ones.