Catalyst

by Zoho

Go to console

QuickML FAQ

Browse through FAQ related to working the various components and features of the Catalyst QuickML service,
such as Data Connectors and Preprocessing, ML Algorithms and Operations, and Pipeline Builders.

×

General

What is the difference between AutoML and QuickML?

AutoML enables you to easily analyse a set of training data and generate predictive analytics on the dataset without requiring you to be involved in the complex ML training process that involves selecting the right ML algorithms to train the model, preprocessing or profiling the data, or managing the models. Catalyst implements the required model training, and automates the entire process for you.

QuickML on the other hand provides you more control in managing ML and data operations, and lets you build, test, deploy, and monitor effective ML models end-to-end. You will be able to perform a host of data preprocessing and transformation operations, pick the ML algorithms for training, and design the pipeline exactly as you need, all with no coding involved.

How safe is my data in QuickML? Will the data be used to increase the QuickML platform's quality?
User data is encrypted and stored safely by adhering to all the security measures applied in Zoho. The data uploaded will not be used to improve the accuracy of any of the QuickML Platform’s algorithms. It is only used to train the intended customer model that user configure.

Is it possible to access a dataset in one Catalyst project from another project?
Catalyst Projects are intended to provide a clear isolation between data that are being handled. It is not possible to access the data from one project from another.

Datasets

How is my data quality score calculated?
Data quality score is calculated by internal metrics considering the invalid and missing values present in the dataset.

What is the maximum number of records that will be fetched from Zoho CRM?
By using Zoho CRM Bulk APIs, maximum of 6 Lakhs records will be fetched into QuickML even if synchronization is configured.

Is it possible pause and resume the synchronization for particular Dataset import?
You can edit the synchronization option of Particular dataset by choosing None option in Sync Frequency Dropdown box in dataset details page.

Pipelines

Is it possible to use multiple datasets in single pipeline?
Yes. You can use multiple datasets in a single pipeline by using the Add Dataset stage and configuring the required dataset in it.

What is the difference between Data Pipeline and ML Pipeline?
Data pipeline is intended to do data pre-processing on the original dataset and reuse the dataset for future pipeline creations. Hence, the data pipeline will only have data operations while the ML pipeline will contain both data and ML operations.

Why is there a statistics difference in dataset details page profile and pipeline source stage profile?
Profile in Dataset details page represents details for whole data but profile in source stage represents details for sampled data from the original dataset.

Is it possible to change the target column of the ML pipeline?
No. Once the target column of the pipeline is saved, it cannot be changed. We will have to create a new ML pipeline with the new target column to experiment.

Is Zia features in QuicML Pipeline builder is generated using customer data?
Zia features are in-house ML/DL models that are pre-trained using large open-source dataset to solve common use-cases and integrated into builder to enhance data preprocessing capabilities.

Why is my current operation processing for long time?

All the pipeline executions are queued and handled asynchronously inside QuickML and based on demand the execution will take place.

In certain cases, the operation might be costlier in terms of computation, so there may be a delay. However, the status of the pipeline execution will be updated respectively once the execution succeeds or fails.

Models

How are the models created?

Model will be created at the successful execution of the Model pipeline automatically.

Once the models are created, we can view the details associated with that model and pipeline under models module.

Endpoints

What is the difference between development and production environments in QuickML endpoints?

While we create an endpoint for a model, QuickML platform automatically enables an REST API endpoint which is to test and verify the model behavior, and is free to use for 1000 invocations.

After verifying the endpoint has to be published to access it in production environment for production grade integrations and charged as per usage.

What are the modes used to integrate QuickML endpoints to external endpoints?

There are two modes available to integrate QuickML endpoints.

REST API: Zoho OAuth authenticated REST API calls.
QuickML SDKs: External calls can be made using Java, Python and NodeJs Catalyst SDKs.

Last Updated 2023-10-08 10:48:45 +0530 +0530