Google Cloud Machine Learning Engine Reviews

Best things about this platform is that we can work very simply and no one disturbs you nothing !! No ads and hassle free platform ...

Also google support is unbeatable ... Ho ahead with this to give your developer platform freeeee hit

Nothing is there to be disliked .. the only thing is that google should provide live chat for products other than this too.

Otherwise the platform is fantastic ... And no one here can challenge it.

I and my whole team would recommend it to fullest and to everyone ...

The people who want to learn or those who want to create will be experiencing amazing things .

We here at my coMpnay yawa this platform for our in company developer. They lesen and they create the best out of it.

You cantik imagine what we have done with bus great platform ... Almost every thing is best. And no one can beat as always..

I like the amount of storage you receive to store finished training modules. These training modules are usually used over and over but differentiated for every group as needed. I also like the guidance given throughout the wizards to help ensure your finished product is awesome. This program is compatible to many devices for me to use at home and at work. The program also uses my current Google drive if I want to for personal training portfolios and modules. Google drive is easily used with Google Cloud Machine because it allows me to share and collaborate with groups. I can also share with administration as necessary and other district members because they also use the Google platform for professional use.

I disliked the way that the Google cloud machine learning did not have a common area to find previously created modules to use. This made it necessary to design from scratch. Sometimes time constraints made this difficult and time consuming.

I recommend using it in a small group before implementing on a large scale. This ensures that you have worked out many kinks and glitches.

I am solving the problem of more effective collaboration and training. Effective feedback can be given with the use of this software because learners can go through at their own pace. Self paced learning is essential in today's educational world. Many learners are learning differently so therefore the use of self paced training and learning is required. This software allows learners to work from anywhere and on their own time. This allows more learners to become engaged with this software and begin using it for personal and business use.

If one uses TF Estimators, running any tensorflow model in a distributed manner becomes absolutely seamless. Over the last few months, the product has become really stable, reporting more detailed and kept nicely up to date with the latest version of tensorflow.

More examples on how estimators are used (Custom Estimators, Canned etc) would be great. Also, being able to run more than the current limit of 25? models in parallel would be great - we often run into our limti.

Not having to think about the distributed nature of running models is a key win for us. We run simple as well as complicated models on Google Cloud ML Engine. Even when running small models which will easily run on a single machine, it made sense for us to use ML Engine - we are able to run a large number of smal independent models in parallel.

I like the measure of capacity you get to store got done with preparing modules. These preparation modules are generally utilized again and again yet separated for each gathering as required. I likewise like the direction given all through the wizards to help guarantee your completed item is great. This program is good to numerous gadgets for me to use at home and at work. The program likewise utilizes my present Google drive in the event that I need to for individual preparing portfolios and modules. Google drive is effortlessly utilized with Google Cloud Machine since it enables me to share and work together with gatherings. I can likewise impart to organization as essential and other area individuals since they additionally utilize the Google stage for proficient utilize.

I think it's on cloud contrasted with other machine learning stages. So once in a while need to experience extra information administration ventures keeping in mind the end goal to put organization information on cloud.

Have utilized it for a wide range of academic projects. Acquainted with NLP and discourse acknowledgment.

I love the ability of the GC ML engine to be able to abstract the complexities of ML algos and enable users to learn and use ML using apis. That's simplicity is really awesome!

I think that learning Tensorflow is still quite complex. It might help to have learning aids and prompts which point you in the right direction when Tensorflow does not work as expected.

Try Amazon but you won't be able to ignore Tensor flow

Trying to work up a virtual receptionist as a use case. I believe the best benefit is being able to scale up the infrastructure required for compute dramatically without any performance challenges.

I like that Google Cloud takes care of ML End to End process flow, I can use buckets to store, tools for monitoring data quality, transform data and spin off TPUs and GPUs as needed for complex algorithms.

Cloud Costs add up; Learning curve for ML/AI scientists to set up Dataflow, IOT, Pubsub etc. Python is not supported everywhere needing Java usage.

Business problems in Agriculture, Medicine, Supply Chain Optimization, Health care

Benefits: Better Monitoring of farming resources via processing satellite imagery, streamline Supply Chain operations and reduce backlog effectively, Generative design models for drugs, material science products, Automated Patient monitoring to save lives.

The best of google cloud machine learning is it’s speed. I get minimal latency in my work compared to other data lake option we had in past. Also, I like the way google cloud provides an option to stich my company data with google adwords and double click data. Plus automated machine learning provided by google and tensorflow is additional advantages.

I find google cloud ML a decent product but the only disadvantage I feel is it’s on cloud compared to other machine learning platforms. So sometimes have to go through additional data governance steps in order to put company data on cloud.

As a data scientist team we need to build churn models and propensity model. Google ML engine helps in building those models and also understanding current customers issues using NLP on their complaints.

i like that i can use ml training and serving as a service. I can programatically set up a series of training jobs

the lack of documentation. you have to guess the concept of operation, constantly go to the help in the command line to explore what is available..... the packaging of your code is annoyingh to do the first two times since you have to guess a lot by following examples. All this could be documented better. Seems like a tool build for cloud/software engineerings trying to do ML than for ML scientist and practitioners.

solution as scale is the desired benefit. Building inference systems related to industrial systems.

Using scikit-learn with pandas and numpy is one way to do ML. But to get the efficiency of distributed training, ML Engine is a must have.

Don't have a easy to follow tutorial to learn different aspect of ML Engine. For example, how to use ML Engine to run Hyper-parameter search, how to do data-parallel training or model-parallel training.

Hyper-parameter tuning and data parallel training.

Google Cloud ML Engine is an awsome product from the Google Cloud Platform that let's you perform serverless machine learning training of models. It will also automatically optimize your models and last but not least let's you deploy your model in production in minutes.

It still requires some coding ability but I guess that's normal we are still talking about machine learning.

Prediction. optimization.

The ease of describing the structure of my neural network and the ease of training it.

I canno think of anything I dislike at the moment. The interface is much cleaner after the introduciton on Keras into this framework.

Attend some training sessions

I am designing a classifier to startify patints into different categories based on their gene expression profiles

Easy to run ML Engine jobs via gcloud ml engine commands

ML models are developed using paths to local files. Updating models to use Google Cloud Storage is difficult and there is no documentation on Google Cloud on how to do this. Also updating python scripts to use argparse is trial and error with little documentation.

Model for case routing

Very easy to sign up and use

REST API and different language sdks

It would be nice if they offered a studio like Azure ML studio to build data-pipelines visually

Image analysis (i.e. multi-object, OCR)

The accuracy and performance of Google Cloud ML services are amazing

I used the computer vision, natural language processing engine. The API support is pretty well that I can write code quickly.

Functions like Safe Search could be improved with more detailed information, instead of justing giving 0-5 ratings.

Image rating with possible texts. The vision problem and text problem could be both solved by ML engine.

Ability to quickly deploy out machine learning without the setup and operational requirements needed to complete the task. Wish there was more ux involved in the machine learning engine but I'm sure that will come.

Requires some development skills and analytical skills and there isn't a simple button you can press.

Sentiment analysis of customers

Being able to do everything in a seamless serverless fashion.

Not everything is tied neatly. Like Firebase data cannot be streamed to Cloud Pub/Sub so we can do machine learning.

Google Speech API doesnt do transcription for longer audio files from contact center recordings

speech translation

matchmaking

object detection in images

I am just trying to learn ML and registered in the ML boot camp. I am enjoying the lectures and how the boot camp is organized.

I got the the info. on pre-requisites just 12 or so hours before the boot camp was supposed to start.

Fraud, metrics, errors

Google AI, having enormous resources and brilliant people have developed AI that can buffer user data and process it in no time.

What I dislike about Google AI is that they are pretty much monopolized the internet world, luckily they're still dependant on other companies.

Smart functions in HTML5

I like how everything is REST based, which makes it easy to use. Moreover, the pretrained models are really useful, esp when the hard part of doing machine learning is getting training data.

I don't like the pricing structure. For things like audio transcription and image recognition, it is fairly pricey, and there are no options to self-host the models (which is bad for hipaa complaince).

Have used it for a variety of different side projects and in industry. Familiar with NLP and speech recognition.

We went through an alpha with GCP particularly for AutoML. The onboarding and commitment of the Google Cloud team is an amazing experience. They seem to have to most experience on this topic compared to the other big competitors like AWS.

The amount of categories is currently fairly limited, but this should expand anytime soon.

We are exploring the possibilities of using AutoML to deliver more insights and intelligence to our CRM offering. We are in an early phase with this.

The Ease of use of GCP ML API. It has the great ability to manage the model management , versioning and ease of deployment . The documentation is very user friendly and it support different languages

The product doesn't provide KPI meterics or the epoc time cycle as well as comparison of all deployed model for better ascore

Healthcare domain. Ability to identify the hospital readmission of patients. Through this API, hospital can serve better patient engagement and better outcomes.

Google Cloud Machine Learning Engine has lots of build in functions in tensor flow and has datalab to run step by step.

Sometimes it takes some time to go thruogh the help ducment.

We are doing project for southern data scientist conference

the whole framework is in place to start with. the data collection api is good. the process to sample the data and create subset is clear. like the idea of data training and prediction. bootcamp is a good start and google cloud platform is worthy registration.

would like it better to be able to register the bootcamp at home and start a free session that valid for a few days.

don't have to set up the whole environment manually.

The open source TensorFlow library. Detailed tutorials that utilize pre-trained models to train specialized vision models. The model zoo.

The console can be improved with better integration with tensorboard and jupyter notebooks.

Automating vision tasks such as object detection.

It is a managed service which allows me to focus on code.

instructions could be simpler. I find it cumbersome to setup the model as python package.

My ability to prototype has improved dramatically.

I like that it's serverless and runs deep learning jobs fairly fast. Easy for already made estimators and image processing where there is lots of examples.

It's pretty hard to figure out if you are new to it. The documentation is hard to find/ or there is lack of documentation. Hard for custom estimators to figure out debugging and tensorboard also deploying the model for me was very hard.

churn prediction

The abstraction of underlying hardware and pay for what you use.

There are model evalauations and other features available on TFX which is not available in ML Engine

Forecasting and new item forecat

I like that the instructors are knowledgeable, and that the training was well organized.

I would like to see more of what's happening under the hood with some of the functions. For example, in the vocabulary_to_feat_cols() function (or similar), it would be nice to understand what this actually does.

GCP Consulting projects as part of PSO. Benefits include: Hyperparameter tuning, storing different model versions, and visualizing training results in TensorBoard.

it's quite intuitive to use and easy once you have a set up. would definitely recommend

there is a bit of a learning curve at the beginning

data analytics with big data

I really like the opportunity to train complex models without buying any hardware in our copany, we can use very super machines and just paying the time that with use them

Somethings ML engine is not working properly

Basically, we are training natural language processing models using Glove and Word2vec because NLP google API is very limited so far, we hope this API can be use in customer entities.

Everything, it has everything that you need to get the job done

Nothing really, the formatting could use tweaking.

It allows me to analyze my problems and recommend things to colleagues

really love GCML. very easy to setup and use.

has a little learning curve to get up and running

being able to get up and running with Machine Learning

GCP machine Learning provides end to end support for a developer.

Machine learning platform needs to have some more developer tools for edge computing

Probems at the edge

I liked that the Google Cloud Learning Engine was easy to pick up.

The only thing that I disliked was actually getting the software to run on my computer.

Definitely give it a go because its easy to pick up.

Right now we are trying to make trains learn through machine learning. We haven't experienced any benefits as of yet but we are working towards it.

easy to use in cloud environment supporting a lot of machine learning algorithms

Not much capability and multiple GPU support

Computer Vision problem like object detection

The products, interfaces and interaction is easy, simple and quick to get a grasp

Nothing so far based on my observation of usage

Employee personalization

The on-click provisioning for the labs is working out nicely

usibility of the tools, some botton is not easy to find

deep learning

The ease with which developers can get started with advanced tools and capabilities is impressive.

I am looking forward to the capabilities Google Contact Center will provide when it's opened up to developers (and not just specific partners).

Support chat bots enhanced with Natural Language processing, to reduce unnecessary escalations to advanced teams.

very user friendly and intutive. can take any large set of data and churn the algorithm within in very less time

Documentation needs to be updated. There is not enough documentation around to point out to right page where my question would be addressed.

Predict adverse events once it is submitted by patients,HCP

Tensorflow allows me to expirement, test with larger data and then deplow. Google Cloud ML enables those efforts really well.

Tensorflow takes a little effort when compared to other options but GCP makes it worth the effort.

We are exploring customer purchase behaviors and how weather impacts in store traffic.

I like that the hypertune features to save my time to search the hyperparameters for the best model. Distributed training is also great.

It is not very user friendly, and does not have a GUI to submit my training job.

train and serve TensorFlow model without maintaining VMs.

Not a data scientist but Google Cloud ML makes it easy to develop and train models. Pleased to hear that it now supports SciKit Learn and XGBoost.

More tutorials and examples would be great!

Using it to try to find trends in patient care and patient satisfaction.

I love how simple it is to connect to my data sources to import models, data sources and logging.

Nothing really -- it helps to have someone walk you through it the first time.

We're trying to spot trends among user requests and predicting problems.

I use datalab and the course from coursera on how to use it. my time has been limited but i enjoy it.

hard to find people who use it. stackoverflow support and slack channel is not that helpful

trying to solve audio classification

Ease to train a model using the data imported from data prep.

Latest Google tensorflow version not often supported on cloud ML.

Going great

Trying to build a human pattern learning system. Its easy, but keeps on upgrading day by day.

Extremely easy and efficient for building models in the cloud without having to worry about infrastructure

Takes time to learn and get started compared to competitor offerings

Increase our capabilities and train models that would not be possible on-premise

The UI and a really easally to understand Frontend is great!

There should be more description to get in touch with the theoretical informatics behind ML.

We use it for HR pruposes, statistics about customers and telecommunication things.

Love its ability to morph its behavior based on my input.

not very well known. to ease the learning curve for me and my clients, i would like to have out-of-the-box solutions to apply.

try it out

Search

Google cloud, auto ml, easy, fast, secure quick trun around

getting the solution more accurateand easiest way . seamlessly.

Nothing as such. All arebest and keep inventing .nothing to complain.

tensor flow

its under evaluation, not been implemented.

Ease of using, and the documentation provided. Reliability from the brand name is assuring too.

The fear data being saved on google servers

Improving tasks like speech recognition with actually having to code for it

Easy to use UI

Good start document in multiple programming languages

Model training time is long. It takes hours to train an small dataset classification

Evaluate multiple vendors for comparison

News type classification

speed and end-to-end. It stramlined the complicated progress, awosome

UI and it is less intutive for me and I need to take time to figure it out

Model for people you may know

hyperparameter tuning and great tutorials

I wish tensorboard was more intuitive to configure

I am trying to develop a credit risk predictor based on checking account data. Tensorflow is the way to go

Good overall knowledge sharing. Good speakers and contnt

Lack of real world useful use cases. I like more real world case studies. Toy example don't serve much purpose

Recommedation, fraud detection.

Tensorflow algorithms are useful when running the machine learning usecases

I have not used extensively so not sure about he dislike

not used it extensively to determine the right use cases

Very powerful and intuitive. Love the support from the product team as well.

Integration with tf.nightly is tricky. Serving is also not straightforward with tf.estimator

Image super resolution. Scalability and power.

Lot of algos, models to choose from, easy setup and management

Not to my knowledge, I am still evaluating it with nothing to compare

Better engagement and retention understanding

All libraries are ready for install and use. No need to worry about update and version

A bit hard to get used to in terms of self learning

Develop

The finding patterns feature is very helpful and accurate

Glitchy/slow sometimes as is every big software application

Software analysis/debugging. Pattern recognition and prediction capabilities are great for application development.

Ready to go models for training. It is a Awesome Engine.

Small time for processing:) It helps everyone to design NN models.

Sure..Its a good product

Vision, Image Recognition

machine learning and deep learning products and features

No so far I dont see anything dislike for the features I am using

still learning it

ability to easily build production ready models

nothing to dislike about this product. This a perfectly good setup

To do nlp for concept extraction

Easy to use. Integrated with TensorFlow and other services.

Not intuitive enoiugh and the UI is crowded and confusing.

NLP. Data anylise is easy.

It is simple and easy to use. Very accurate.

Authentication sometimes doesn't work in one go.

Integrated Translation and Transcription into our services.

Tensorflow and the pipeline thst work seamlessly

The support and the user experience has room for improvement

Ml on the edge

flexibility and speed in developing applications

like to see multi tenant solutions that can scale

nlp