Interview with Google Programmer Martin Horner about TensorFlow
Continuing a series of interviews with speakers PyCon Russia conversation with Martin Gorner (Paris, France).
Martin Gorner was at the origin of e-books since the launch of Mobipocket, which later became part of the software on Amazon Kindle and its mobile versions, and since 2011 Martin has been working at Google, where he is actively engaged in machine learning and TensorFlow A new, fast, smart and flexible machine learning system that can work on a simple smartphone, as well as on thousands of nodes in data centers.
Below is a short interview with Martin about what TensorFlow is, why Google opened TensorFlow for developers in open source, and how the system may be of interest to developers who are not familiar with machine learning.
')
- Could you please tell me a bit about TensorFlow? What is this: a set of low-level libraries or a service or a framework?
- Tensorflow is a framework for distributed computing and hardware-accelerated scientific computations. It can be expressed in terms of matrix multiplications and needs to iterate and converge. It uses a model of computations and then execute it. It is a graph of multiple GPUs. For example, your computed nodes. Tensorflow handles these kinds of low-level logistics.
- Would you like to be acquainted with Machine Learning much? How can it be useful for them?
- Tensorflow today has a high-level library called tensorflow.nn specifically designed for neural networks. We will be expanding. Also, the Google Cloud service will be optimized for the neural network training and serving. So today Tensorflow is very specialized in neural networks. That's it. I) I hope I’m looking for it. 2) Tensorflow is such a tool.
- Google doesn’t share it's developments with the OpenSource community very often. Sometimes it’s what Google’s internal services have. Why did Google decide to open source TensorFlow?
- This is a general trend. It is not just the papers. The Apache Beam project is another such example. The “dataflow model” has been published as a paper but also open-sourced as Apache Beam. This way, you can use data flow, you can use Google Cloud Dataflow. Kubernetes (Google's container deployment and management technology) followed the same path too. It has now been embraced by IBM, Microsoft, VMWare and others.
- What is Google's approach to management of TensorFlow's development?
- This is also heavily used internally at Google. Actually, it is the second iteration internally. We had a first version called DistBelief from which we learned a lot and restarted from scratch. For our customers, we have been able to complete our project. The community response has been fantastic. Tensorflow became the most-forked machine learning project almost overnight. Since then, the community contributions have been welcome. Tensorflow TFLearn is a higher-level learning program for Tensorflow.
- How big is the contribution to TensorFlow from third-party developers? Was it useful for google so far?
- Sourcing is not necessary for Google. We will be contributing to this field. It is a process of learning how to use. The foundational bridge between research and real-world applications. Good developers, even if they’re not being specialized in machine learning, they’re given the right tools. It is a way of life for women.
On July 3rd, Martin will speak at PyConRu - talk in detail about TensorFlow and hold a workshop for those who want to get a deeper look into it. There will be an opportunity to meet Martin personally.
Join now!
Thanks to our sponsors: General Sponsor - Positive Technologies , Gold Sponsor - JetBrains , Silver Sponsors - Rambler & Co and Wargaming , Bronze Sponsors - Ostrovok.ru and Scrapinghub .
Martin Gorner was at the origin of e-books since the launch of Mobipocket, which later became part of the software on Amazon Kindle and its mobile versions, and since 2011 Martin has been working at Google, where he is actively engaged in machine learning and TensorFlow A new, fast, smart and flexible machine learning system that can work on a simple smartphone, as well as on thousands of nodes in data centers.
Below is a short interview with Martin about what TensorFlow is, why Google opened TensorFlow for developers in open source, and how the system may be of interest to developers who are not familiar with machine learning.
')
On July 3-4, Martin will be at the conference PyCon Russia 2016
- Could you please tell me a bit about TensorFlow? What is this: a set of low-level libraries or a service or a framework?
- Tensorflow is a framework for distributed computing and hardware-accelerated scientific computations. It can be expressed in terms of matrix multiplications and needs to iterate and converge. It uses a model of computations and then execute it. It is a graph of multiple GPUs. For example, your computed nodes. Tensorflow handles these kinds of low-level logistics.
- Would you like to be acquainted with Machine Learning much? How can it be useful for them?
- Tensorflow today has a high-level library called tensorflow.nn specifically designed for neural networks. We will be expanding. Also, the Google Cloud service will be optimized for the neural network training and serving. So today Tensorflow is very specialized in neural networks. That's it. I) I hope I’m looking for it. 2) Tensorflow is such a tool.
- Google doesn’t share it's developments with the OpenSource community very often. Sometimes it’s what Google’s internal services have. Why did Google decide to open source TensorFlow?
- This is a general trend. It is not just the papers. The Apache Beam project is another such example. The “dataflow model” has been published as a paper but also open-sourced as Apache Beam. This way, you can use data flow, you can use Google Cloud Dataflow. Kubernetes (Google's container deployment and management technology) followed the same path too. It has now been embraced by IBM, Microsoft, VMWare and others.
- What is Google's approach to management of TensorFlow's development?
- This is also heavily used internally at Google. Actually, it is the second iteration internally. We had a first version called DistBelief from which we learned a lot and restarted from scratch. For our customers, we have been able to complete our project. The community response has been fantastic. Tensorflow became the most-forked machine learning project almost overnight. Since then, the community contributions have been welcome. Tensorflow TFLearn is a higher-level learning program for Tensorflow.
- How big is the contribution to TensorFlow from third-party developers? Was it useful for google so far?
- Sourcing is not necessary for Google. We will be contributing to this field. It is a process of learning how to use. The foundational bridge between research and real-world applications. Good developers, even if they’re not being specialized in machine learning, they’re given the right tools. It is a way of life for women.
On July 3rd, Martin will speak at PyConRu - talk in detail about TensorFlow and hold a workshop for those who want to get a deeper look into it. There will be an opportunity to meet Martin personally.
Martin's report is called “TensorFlow and deep learning, without a PhD”
Join now!
Thanks to our sponsors: General Sponsor - Positive Technologies , Gold Sponsor - JetBrains , Silver Sponsors - Rambler & Co and Wargaming , Bronze Sponsors - Ostrovok.ru and Scrapinghub .
Source: https://habr.com/ru/post/303832/
All Articles