Four no-code deep learning platforms under the microscope

Already very present in AI, no-code development is investing in the field of artificial neural networks. Overview of some tools positioned in this area.

Artificial intelligence hasn’t escaped the no-code wave, with development tools multiplying in recent years. The challenge ? Putting machine learning previously reserved for data scientists within the reach of data analysts and other business professionals. Predictive analytics, bot platform, text recognition… In many areas of AI, no code prevails. He also invests in deep learning and deep neural networks. The JDN has selected four solutions positioned in this area.

Apple Create ML: one tool for everyone

Create ML is a no-code machine learning solution developed by Apple for macOS users. Its peculiarity is that it appeals to the general public. You don’t have to be a developer or data scientist to use it. It enables neural networks to be trained to recognize images, sounds or text. Built on the same technology as the Siri or Photo applications, Create ML is based on Apple’s machine learning framework: Core ML. It is equipped with a graphical console designed to evaluate the performance of the learned models. To carry out the learning process, Apple provides graphics processing power (GPU) in cloud mode if required. Ultimately, the models can be used on macOS, but also on iOS, iPadOS and tvOS.

Create ML is aimed at the general public. © JDN / Capture

H2O Hydrogen Torch: a tool for Citizen Data Scientists

H2O Hydrogen Torch, published by the American H2O, is a no-code platform solely focused on the development and deployment of artificial neural networks. Under the hood is a variety of open-source reference technologies: PyTorch, TorchVision, MLFlow, Scikit-Learn, etc. “Beginners can get started quickly by selecting only the most important parameters of a network, while experts can enable a wide range of learning methods ‘ explains H2O. From end to end, the user remains in control of the machine learning process and parameter adjustment. The models are then automatically packaged to go into production. Targeted Use Cases? Automatic speech processing and image recognition.

H2O Hydrogen Torch is aimed at both novice data scientists and experienced data scientists. © JDN / Capture

Obviously AI: an auto ML tool

Obviously, AI is presenting its offering as an automated machine learning cloud platform. This San Francisco company doesn’t hesitate to add the word “no code” to its solution. From a given dataset (e.g. a .csv file), the application automatically selects an algorithm to be adapted to the desired prediction: regression model, decision tree, random forest… In addition to statistical algorithms, this can of course also be used in AI Think of a neural network, especially when it comes to recognizing images or managing NLP (natural language processing) tasks. Finally, Obviously AI allows sharing the model via a simple link or API.

Obviously, AI is presenting its offering as an automated machine learning cloud platform. © JDN / Capture

Clarifai: a tool for computer vision

Clarifai specializes in image recognition and articulates its offering, like Obvious AI, around an automated machine learning platform. Its most important added value? A library of pre-trained models. These cover various computer vision topics: logo recognition, face recognition, vehicles, clothing, etc. They also extend to NLP, from simple text classification to sentiment analysis. “These pre-trained models make it possible to identify biases and unbalanced samples due to oversampling or the integration of data contrary to regulations,” points out one from Clarifai. “For example, our demographic models identify age, gender, and multicultural makeup to help you identify imbalances in these different areas.”

Clarifai comes with a library of pre-trained models. © JDN / Capture

methodology : The solutions presented above were selected by the JDN editors. This is not an exhaustive presentation of the no-code deep learning tools on the market.

Leave a Reply

Your email address will not be published.