Baidu open-sources Paddle Quantum toolkit for AI quantum computing research

Baidu today announced Paddle Quantum, an open source machine learning toolkit designed to help data scientists train and develop AI within quantum computing applications. It’s built atop Baidu’s PaddlePaddle deep learning platform, and Baidu claims it’s “more flexible” compared with other quantum computing suites, reducing the complexity of one popular algorithm — quantum approximate optimization algorithm (QAOA) — by a claimed 50%.

Experts believe that quantum computing, which at a high level entails the use of quantum-mechanical phenomena like superposition and entanglement to perform computation, could one day accelerate AI workloads compared with classical computers. Moreover, AI continues to play a role in cutting-edge quantum computing research. Baidu is aiming to target researchers on both sides of the equation with Paddle Quantum — toolkits that include quantum development resources, optimizers, and quantum chemistry libraries.

Paddle Quantum supports three quantum applications — quantum machine learning, quantum chemical simulation, and quantum combinatorial optimization — and developers can use it to build quantum models from scratch or by following step-by-step instructions. It includes resources addressing challenges like combinatorial optimization problems and quantum chemistry simulations, as well as complex variable definitions and matrix multiplications enabling quantum circuit models and general quantum computing. It also features an implementation of QAOA that translates into a quantum neural network by identifying a model through classical simulation or running directly on a quantum computer.

“Since Baidu announced the establishment of [the] Institute for Quantum Computing in March 2018, one of our primary goals [has been] to build bridges between quantum computing and AI,” Baidu said in a statement. “[Paddle Quantum] … can help scientists and developers quickly build and train quantum neural network models and provide advanced quantum computing applications.”

Today Baidu also unveiled the latest version of its PaddlePaddle machine learning framework, which over the past few months has gained 39 new algorithms for a total of 146 and more than 200 pretrained models. Among them are Paddle.js, a deep learning JavaScript library that allows developers to embed AI within web browsers or programs in apps like Baidu App and WeChat; Parakeet, a text-to-speech toolkit with cutting-edge algorithms like Baidu’s latest proposed WaveNet model; Paddle Large Scale Classification Tools (PLSC), which enables image classification model training across graphics cards; and EasyData, a new drag-and-drop data service for data collection, labeling, cleaning, and enhancement.

Over 1.9 million developers now use PaddlePaddle, according to Baidu, and 84,000 enterprises have created more than 230,000 models with the framework since its debut — up from 65,000 enterprises and 169,000 models as of last November. (PaddlePaddle, which was originally developed by Baidu scientists for the purpose of applying AI to products internally, was open-sourced in September 2016.) The company anticipates growth will accelerate in light of the recently relaunched PaddlePaddle hardware ecosystem initiative, which will see manufacturers such as Intel, Nvidia, Arm China, Huawei, MediaTek, Cambricon, Inspur, and Graphcore contribute expertise and promote AI app development.

The unveiling of Paddle Quantum follows the release earlier this year of Google’s TensorFlow Quantum, a machine learning framework that can construct quantum data sets, prototype hybrid quantum and classic machine learning models, support quantum circuit simulators, and train discriminative and generative quantum models. Facebook’s PyTorch has its own multi-contributor project for quantum computing in PennyLane, a library for quantum machine learning, automatic differentiation, and optimization of hybrid quantum-classical computations. And Microsoft offers several kits and libraries for quantum machine learning applications.