Top 5 Machine Learning-as-a-Service providers
Machine learning is the next big thing in computing; are you ready for it? Hiring data scientists or ML experts isn’t easy or cheap. But the rise of machine learning-as-a-service (MLaaS) suggests that you won’t need to. Today, we take a look at five of the top machine learning service providers to see which one works the best for you.
However, hiring machine learning experts is something of an issue, as the demand continues to outstrip the supply. What’s more, hiring a ML expert isn’t cheap, as they regularly command some of the highest salaries in tech.
Enter machine learning as a service (MLaaS). Much like other –aaS offers, MLaaS provides users with a range of tools as part of a cloud computing service. This can include tools for data visualization, facial recognition, natural language processing, image recognition, predictive analytics, and deep learning.
In the case of MLaaS, the provider handles the actual computations in their own data centers. The customers do not have to install their own software or run their own servers. Generally, the first hit is free with ML services for developers so they can evaluate a platform’s usefulness before subscribing.
MLaaS is a pretty good midway point for companies that want to dip their toes into the machine learning trend without diving right in. Having an established provider is an excellent way to help minimize any transitional issues and gives a certain surety to the whole proceedings. In any case, most of the big names in tech have their own MLaaS platform service.
Let’s take a closer look, shall we? In no particular order:
Microsoft Azure has a whole host of services available for the developer in need. But their machine learning offerings are particularly useful. Azure boasts scalable machine learning, for all sizes. They’re suitable for AI beginners and experts alike, with a range of tools that tend to be more flexible for out-of-the-box algorithms.
The main MLaaS from Microsoft Azure is the ML Studio. It has something of a steep learning curve, since almost all operations must be completed manually, from data exploration, preprocessing, choosing methods, and validating modeling results. However, to make things easier, the browser-based environment is highly simplified with a visual drag-and-drop mechanism. No coding is necessary here!
ML Studio has a huge variety of algorithms at your disposal, with around 100 methods for developers to play with. Additionally, the Cortana Intelligence Gallery is a community based collection of ML solutions used by data scientists.
ML Studio’s most popular option is the free workspace, which only requires a Microsoft account of some kind. This includes free access that never expires, 10GB of storage, R and Python support, and predictive web services. The standard enterprise grade workspace is a little pricier, with $9.90/month and a Azure subscription, but it has a lot more support and services available.
Amazon Web Services have more or less revolutionized the SaaS field. It’s no surprise that they are also a dominant player in the MLaaS department as well. Amazon Machine Learning is an incredibly popular service that guides users through creating ML models without needing to learn the complex algorithms themselves. Once you’ve created you models with the visualization tools and wizards, simple APIs create predictions for your application without any need for generating code or managing infrastructure.
There’s a high level of automation available with Amazon Machine Learning, making it useful for beginners. The service can load data from multiple sources, including Amazon RDS, Amazon Redshift, CSV files, and more. The service figures out which fields are categorical and numerical and determines the accurate methods of data preprocessing on its own. However, Amazon ML does not allow any unsupervised learning methods, making it hard for beginners to develop their own understanding.
Pricing for Amazon ML is on a pay-as-you-go model. There is a flat $0.42/hour fee for data analysis and model building, with a separate fee for batch predictions ($0.10 per 1,000 predictions, rounded up to the next 1,000) and real-time predictions ($0.0001 per prediction, rounded up to the nearest penny). Data storage is billed separately. So, your final bill depends on how fast and efficient you are.
We covered Watson Machine Learning when it first debuted for a general audience last year. WML is a general service provider that runs on IBM’s Bluemix. Capable of both training and scoring, WML is designed to help both data scientists and developers.
WML is intended to address questions of deployment, operationalization, and even deriving business value from machine learning models. Users can keep utilizing their own Jupyter notebooks in Python, R, and Scala. Watson ML also boasts visual modelling tools that help users quickly identify patterns, gain insights, and make decisions faster. Check out our review to see some of the nifty tools available with Watson Machine Learning.
This is a major part of IBM’s Watson Data Platform, and pricing does tend to reflect that. You will need to create an account with Bluemix to start playing around with the service, but there’s a 30 day free trial and it’s pretty fun. After that, you need to choose between Lite, Standard, and Professional. While Lite is free as along as you stay under 5,000 predictions and 5 compute hours, Standard and Professional depend on how intensely you need your computing hours. Predictions run $0.40 – $0.50 per 1000 predictions.
Google’s range of SaaS is nearly endless. So, it’s no surprise that they also have a cutting edge MLaaS platform available. Across their Cloud AI services, there’s a Machine Learning Engine, as well as services for natural language processing and APIs for speech, natural language processing, translation, video, and image recognition. We also covered Cloud AutoML earlier this year here.
However, it’s Google’s Cloud Machine Learning Engine that we’re looking at today, which offers users an easy alternative to build ML models for data of any type or size. Data scientists should be excited to note that Google ML Engine is highly flexible and based off the ever-popular TensorFlow project. Of course, this platform is integrated with all the other Google services, but it’s mostly aimed at deep neural network tasks.
As for cost, if you’re interested in trying out Cloud ML Engine, you can sign up for a free trial here. There’s no initial charge and it comes with a $300 credit. However, a subscription to Google Cloud Platform isn’t cheap.
BigML is something of the odd service out: it’s the only MLaaS provider on this list that isn’t backed by a tech giant. Nonetheless, your options are hardly limited with BigML. While it is platform-agnostic, BigML allows data imports from all possible options from AWS, MS Azure, Google Storage, Google Drive, Dropbox, and more. This kind of cross-provider access is rare as far as MLaaS is concerned.
BigML’s focus on machine learning means that it has more features available that are integrated into its web UI. The platform is intuitive and easy to use, with a flexible deployment for any enterprise option. BigML also boasts a large gallery of free datasets and models to play with, useful clustering algorithms and visualization, anomaly detection, and more.
BigML has a number of pricing options. If you’re clever with your dataset sizes, you can perform unlimited tasks for datasets up to 16MB for free. Students and educators have discounts as well. There’s even a pay-as-you-go option. For more private deployments, BigML offers options for companies with more stringent data security or privacy requirements.