The Developer Community Tech Week is back with you, so check out some of the week’s most important news for our developers.

  • Digital currency 50,000 200 red envelopes, JD.com appointment draw!
  • Cloud data: Forecast of four major trends in the artificial intelligence data collection industry in 2021
  • Google drives will integrate enterprise and personal services and incorporate desktop client software
  • Microsoft released Win10 cloud configuration: flexible deployment of Win32 applications and other security
  • Faster and clearer! Nvidia implements SDF real-time rendering for the first time, increasing speed by three orders of magnitude
  • Nvidia A100 deep learning performance test: training speed up to 3.5 times of V100
  • AAAI 2021 Tsinghua proposed a deep pair of oligomerization classes for new intention discovery
  • ACM MM 3D Dense Connection Network Based on Time-frequency-space Characteristics of Emotional EEG Signals

Industry News

1. Digital goods 50,000 200 red envelopes, JD.com appointment draw!

Digital RMB is coming to Beijing! Open the “Jingdong” APP, search for “digital RMB”, from February 7 to February 8, Beijing area can sign up for a lottery (based on GPS positioning) to draw 200 yuan digital currency red envelope, limited to 50,000. The result of the lottery will be announced on February 10. From February 10 to February 17, you can use red envelopes, which can be spent online at Jingtokyoxi. If the lottery is won, JD.com will send the winning information via text message.

Supported banks: ICBC, Agricultural Bank of China, Bank of China, Construction Bank of China, Bank of Communications, Postal Savings.

2. Forecast of the four major trends in the AI data collection industry in 2021

Artificial intelligence data collection and annotation head enterprise cloud survey data recently released the “forecast of the four major trends in the artificial intelligence data collection and annotation industry in 2021”. Cloud data believes that in the future, high-precision data will become a hot spot in the AI training stage, with the characteristics of stable main demand side, and the stock market will grow steadily. With the further expansion of artificial intelligence’s demand for long-tail scene data, scenario-based data will have a broader incremental space in 3-5 years and become a new engine for the accelerated development of the industry. At the same time, the “underlying technology + service capability” will become more and more important until it becomes the core competition point, and the service providers that can provide integrated data solutions are more needed for AI.

3. Google drives will integrate enterprise and personal services and incorporate desktop client software

According to Neowin, Google plans to merge its corporate and personal GHDD services to solve the problem of user confusion. The Google cloud Drive (Google Drive) currently has two desktop clients, one for enterprise users and one for individual users. Google announced yesterday that it will integrate the two services later this year, with the aim of making it easier for managers to synchronise file distribution and to avoid confusion among employees. In addition, Google will also design a new Google Drive desktop client to unify the two services.

4. Microsoft released Win10 cloud configuration: flexible and safe deployment of Win32 applications

February 3 – Microsoft today announced the release of a cloud-based version of Windows10, according to its official blog. Cloud Config provides a simple way to apply a uniform set of configurations to Windows 10 devices, turning them into streamlined, easy-to-manage terminals. To be clear, cloud configuration is not a new version of Windows or a new model. It is simply a Windows 10 device configuration recommended by Microsoft, with cloud optimization for users with specific workflow needs. With Windows 10 cloud configuration, end users will get a familiar Windows interface and applications, while IT administrators will get a familiar administrative experience. If you have a Windows 10 Pro or Enterprise PC and Endpoint Manager, you can use the cloud to configure applications and Settings. In addition to Microsoft Teams, Microsoft Edge, and OneDrive for Business, cloud configurations give you the flexibility to securely deploy Win32 and line-of-business (LOB) applications to devices, either directly or through virtualization.

5. Faster and clearer! Nvidia implements SDF real-time rendering for the first time, increasing speed by three orders of magnitude

Nvidia recently published research that increased real-time rendering speeds by 2-3 orders of magnitude. In terms of rendering quality, it is also better able to handle complex styles, proportions of graphics data, and even shadows that may be generated by ambient lighting in real time. The latest study, titled “Level of Neurogeometric Detail: Real-time Renderings of Implicit 3D Shape,” was published by Nvidia in collaboration with researchers from the University of Toronto and McGill University and is now available on the preprint paper repository arXiv.

In the paper, the researchers say they achieved the first SDF-based 3D high-fidelity real-time rendering while achieving state-of-the-art geometric reconstruction quality by introducing an efficient neural network representation method. More importantly, it improved the rendering speed by 2-3 orders of magnitude compared to other studies.

6. Nvidia A100 deep learning performance measurement: training speed can reach 3.5 times of V100

Nvidia recently released the A100 Accelerated Computing Card based on the new architecture, the website says: automatic mixing accuracy and FP16, can deliver a double performance improvement; Moreover, the A100 with the TF32 can perform 20 times better than the Nvidia Volta without changing the code. Compared with the V100, the A100’s single-precision floating-point computing capacity has increased from 15.7TFLOPS to 19.5TFLOPS. Double-precision floating-point computing increased from 7.8TFLOPS to 9.7TFLOPS.

Follow the Academic News

1.AAAI 2021 Tsinghua University proposed a deep pair of oligomerization classes for new intention discovery

The DeepAligned algorithm successfully transfers knowledge with a small number of known intent priors and learns high-quality intent feature representations for new intent discovery through deep pairs of oligometerization classes. The method can predict the number of new intents by using a small number of labeled intents samples and provide consistent pseudo-labels as self-supervised signals for representation learning through alignment policies. We experimented on two benchmark datasets, and DeepAligned provides a significant improvement over existing approaches with different known intents and cluster number Settings.

* link: https://arxiv.org/pdf/2012.08…

2.ACM MM 3D dense connection network based on time-frequency-space characteristics of emotional EEG signals

In this paper, an EEG emotion recognition model, SST-EmotionNet, is proposed. This model captures EEG signals simultaneously using the structure of two streams, which has three complementary characteristics of space, frequency and time. At the same time, a space-frequency-time-attention mechanism is designed to enable the model to dynamically pay attention to some time-frequency-space local patterns that are more identifiable for emotion recognition tasks. Experiments on the publicly available EEG emotional recognition datasets SEED and SED-IV show that this model has excellent performance over the traditional baseline approach. In addition, the proposed model is a general framework of multivariate physiological time series, which can be applied to sleep stages, fatigue driving monitoring and other related fields.

* link: https://dl.acm.org/doi/abs/10…

The above information comes from the network, edited by the public number of “JD Technology Developer”, and does not represent the position of JD Technology

Recommended reading

  • OpenAI extends K8S to 7,500 nodes to support machine learning; Graph Diffusion Network improves the accuracy of traffic flow prediction
  • Changes to Elasticsearch and Kibana open source licenses; Facebook is using AI to enhance the ability to describe photos for the visually impaired
  • The Docker launch team’s top three directions for 2021; The Ministry of Industry and Information Technology issued the Action Plan for Industrial Internet Innovation and Development (2021-2023).

Please click on [JD Technology] to learn more about the developer community

More wonderful technical practice and exclusive dry goods analysis

Welcome to pay attention to the official account of “JD Technology Developer”