More and more software engineer development AI applications. This is a bit different from the traditional software projects, with iterative cycles of requirement analysis, design, coding, testing, demonstrating.
They have to make sure that complex algorithms are embedded in the code, they may be want to get some insight into the model, lot's of (streaming) data has to be collected, etc.
We are interested in your idea's about this.
I would wonder what you mean under AI applications because there are lots of them with different requirements in regard of analysis, design, coding, testing, demonstrating.
It all depends on what domain your AI application could serve in, and what tasks are going to be done.
I think that jupyter is good enough in buliding the system during the implementation phase.
Dear Gerard Schouten
The best software package for AI application development with complete software development cycle is visual studio see the following two sites:
Install Visual Studio Tools for AI
https://docs.microsoft.com/en-us/visualstudio/ai/installation?view=vs-2017
Improving the Software Development Lifecycle with Visual Studio
https://www.slideshare.net/matthewphillips/improving-the-software-development-lifecycle-with-visual-studio-team-system
I would suggest that you must have know-how of most of the machine learning algorithms and the tool which aids the AI developers is mostly Python. However, you can try the following based on your requirements:
https://developers.google.com/ml-kit/
https://www.tensorflow.org/
https://mxnet.apache.org/
https://www.edgeverve.com/artificial-intelligence/nia/
It realy comes down to the level you are in. On PC-applications we successful used a C# platform. As neuronal networks can be used on Edge (IOT) e.g. for sensor fusion then selecting MCUs with FIR-Filter coprocessors are quite helpful. Such devices are programmed in C or C++ with manufacturer libraries.
there is such a tool: the Eidos system:
http://lc.kubagro.ru/aidos/_Aidos-X.htm
One thing that is often overlooked is the operations/production phase. Many tools for developing algorithms do fine on the desktop but when you need to put them into production (especially at scale) there are a whole new suite of issues to deal with.
So during your development, start considering factors like:
- * What platform do my production systems run on?
- * Can I use a Cloud or do I need to do this on-premises? If a Cloud, can I afford to be bound to one provider (Amazon, Microsoft, Google)?
- * What scale of use is this going to get in terms of number of users/callers?
- * What level of performance do I require? How does it affect accuracy and vice versa?
One thing to look at strongly is if your solution can be containerized with Docker -- it can make these conversations much easier. It can also help your engineers develop faster and with less cries of "it worked on my machine!"
If you went out for a walk, you would be processing millions of bits of information without even being conscious of it. The data would come in: is it raining, is it hot, is it noisy, etc. and you would plan the simplest of things like how to walk, to avoid the gutter building up the rain, to go slow so as not to get over-heated, etc. AI applications will need to have similar sources of data and to learn the parameters that can help to make predictions about how to adjust for the next time. Even you might have walked in the gutter in the rain until you learned that you did not like to have your feet wet.
So, it is not as much about the language in my opinion, it is about how the software can take in a lot of data, analyze the data, understand the patterns, and predict beyond the patterns so that the software can learn to adjust. I would do this in any language you are most comfortable with first since it will be more important to think about the problem than the code.
That depends on the data type you need to treat, and on the domain of data treated. I've searched on the net, and I found this article:
https://digi117.com/blog/top-5-tools-for-artificial-intelligence-ai-and-machine-learning-ml-development.html
In fact, first is necessary to check the type of AI technique and the APP architecture. After that, for AI development use tools as TensorFlow, scikit-learn, java-ml, or build it by yourself.
This link talk about:
Top AI Software Development Tools in 2020
https://levelup.gitconnected.com/top-ai-software-development-tools-in-2020-85618d01b1c4
Here
https://medium.com/@unbug/how-to-build-an-ai-platform-2493d57c9e8
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence