MongoDB vs. Pytorch

If asked by a colleague I would highly recommend MongoDB. MongoDB provides incredible flexibility and is quick and easy to set up. It also provides extensive documentation which is very useful for someone new to the tool. Though I've used it for years and still referenced the docs often. From my experience and the use cases I've worked on, I'd suggest using it anywhere that needs a fast, efficient storage space for non-relational data. If a relational database is needed then another tool would be more apt.

Incentivized

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

• Being a JSON language optimizes the response time of a query, you can directly build a query logic from the same service
• You can install a local, database-based environment rather than the non-relational real-time bases such a firebase does not allow, the local environment is paramount since you can work without relying on the internet.
• Forming collections in Mango is relatively simple, you do not need to know of query to work with it, since it has a simple graphic environment that allows you to manage databases for those who are not experts in console management.

Incentivized

• 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

• An aggregate pipeline can be a bit overwhelming as a newcomer.
• There's still no real concept of joins with references/foreign keys, although the aggregate framework has a feature that is close.
• Database management/dev ops can still be time-consuming if rolling your own deployments. (Thankfully there are plenty of providers like Compose or even MongoDB's own Atlas that helps take care of the nitty-gritty.

Incentivized

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

Incentivized

I am looking forward to increasing our SaaS subscriptions such that I get to experience global replica sets, working in reads from secondaries, and what not. Can't wait to be able to exploit some of the power that the "Big Boys" use MongoDB for.

Incentivized

NoSQL database systems such as MongoDB lack graphical interfaces by default and therefore to improve usability it is necessary to install third-party applications to see more visually the schemas and stored documents. In addition, these tools also allow us to visualize the commands to be executed for each operation.

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

Finding support from local companies can be difficult. There were times when the local company could not find a solution and we reached a solution by getting support globally. If a good local company is found, it will overcome all your problems with its global support.

Incentivized

While the setup and configuration of MongoDB is pretty straight forward, having a vendor that performs automatic backups and scales the cluster automatically is very convenient. If you do not have a system administrator or DBA familiar with MongoDB on hand, it's a very good idea to use a 3rd party vendor that specializes in MongoDB hosting. The value is very well worth it over hosting it yourself since the cost is often reasonable among providers.

Incentivized

We have [measured] the speed in reading/write operations in high load and finally select the winner = MongoDBWe have [not] too much data but in case there will be 10 [times] more we need Cassandra. Cassandra's storage engine provides constant-time writes no matter how big your data set grows. For analytics, MongoDB provides a custom map/reduce implementation; Cassandra provides native Hadoop support.

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

• Open Source w/ reasonable support costs have a direct, positive impact on the ROI (we moved away from large, monolithic, locked in licensing models)
• You do have to balance the necessary level of HA & DR with the number of servers required to scale up and scale out. Servers cost money - so DR & HR doesn't come for free (even though it's built into the architecture of MongoDB

• 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