TensorFlow.js 0.10.3 发布，在浏览器中训练 ML 模型

TensorFlow.js 是一个开源硬件加速 JavaScript 库，用于训练和部署机器学习模型。

TensorFlow.js 0.10.3 已发布，包含大量更新：

Core (0.7.1 ==> 0.8.4)

• Add tf.toPixels
• Add tf.erf op
• Add tf.movingAverage
• Add tf.resizeNearestNeighbor
• Add slice ergonomics
• Finish gradient for tf.pow
• Add gradients for tf.batchNormalization4d
• Finish gradient for tf.matMul
• Add tf.logSigmoid and tf.softplus and gradients

Breaking changes

• Require shape in tensor2d/3d/4d(flatValues, shape)
• Align API with TF
• Align webgl and cpu backend to be closer to TF

Bug fixes

• Implement floordiv for integer division & update CONTRIBUTING.md

Performance

• Make CPU matmul 100x faster

TypeScript

• Modify tf.split signature to preserve rank

Development

• Add yarn “link-local” and yarn “publish-local”
• Add quotemark rule to tslint
• Make core tests reusable by other packages

更多内容和下载地址可查阅发行列表。

转自 https://www.oschina.net/news/95556/tensorflowjs-0-10-3-released