Kortical vs. Pytorch

Kortical is really widely applicable to many use cases, although it doesn't handle images or video it is great to help you build really great ML models without needing to plan ahead what you are going to try, you let the platform build you the best model. It is suited to beginner and more advanced data scientists as you can edit the code to narrow the search space which makes model creation more you build it without AutoML. Hosting the model behind an API that is ready to go is great as it saves so much time vs doing that dev work from scratch

They have created Pytorch Lightening on top of Pytorch to make the life of Data Scientists easy so that they can use complex models they need with just a few lines of code, so it's becoming popular. As compared to TensorFlow(Keras), where we can create custom neural networks by just adding layers, it's slightly complicated in Pytorch.

Incentivized

• The NLP models results were much better than the ones that we did outside of the platform.
• It is really easy and quick to build a good model with a lot of the manual boring tasks all done automatically like one hot encoding, etc.
• Kortical shows the features and their importance for any model type as part of the platform which is great for understanding the models.

• flexibility
• Clean code, close to the algorithm.
• Fast
• Handles GPUs, multiple GPUs on a single machine, CPUs, and Mac.
• Versatile, can work efficiently on text/audio/image/tabular datasets.

Incentivized

• It would be ideal to have Jupyter built into the platform, they say it is coming.
• Also while it is easy to use, at the start it would have been helpful to have more help guides.

• Since pythonic if developing an app with pytorch as backend the response can be substantially slow and support is less compares to Tensorflow

Incentivized

The big advantage of PyTorch is how close it is to the algorithm. Oftentimes, it is easier to read Pytorch code than a given paper directly. I particularly like the object-oriented approach in model definition; it makes things very clean and easy to teach to software engineers.

Incentivized

Their support is great as we use Slack and we have our own channel and they always respond really quickly. Data Science support is available to help unblock you as well as dev support as we're setting up the data feeds. It would be great if there were more FAQ or self-help guides in the platform but the personal touch is also really appreciated and probably gets us there quicker anyway.

Pytorch is very, very simple compared to TensorFlow. Simple to install, less dependency issues, and very small learning curve. TensorFlow is very much optimised for robust deployment but very complicated to train simple models and play around with the loss functions. It needs a lot of juggling around with the documentation. The research community also prefers PyTorch, so it becomes easy to find solutions to most of the problems. Keras is very simple and good for learning ML / DL. But when going deep into research or building some product that requires a lot of tweaks and experimentation, Keras is not suitable for that. May be good for proving some hypotheses but not good for rigorous experimentation with complex models.

Incentivized

• ROI is great as what we would spend on compute we get the AutoML for essentially the same price so it is cost neutral as Kortical comes with compute built-in.
• The results mean that we can automate so much more than our previous model so that is key to the positive ROI.
• The platform auto trains new models and lets us know when there is a better model so it has saved a lot of time so we can focus on new business problems to solve with ML.

• The ability to make models as never before
• Being able to control the bias of models was not done before the arrival of Pytorch in our company

Incentivized