Overview
Neural networks are the status quo of ML today, but relying on abstractions of the original data doesn't allow for full explainability and increases the risk of bias, hallucinations, and model drift. Howso runs alongside AI models and allows for model monitoring and debugging of AI and their data. Howso leverages the data itself, rather than an abstraction of the data, allowing for the elimination risks like bias, drift, and lack of transparency and therefore the full realization of enterprise AI/ML investments. Howso leverages instance-based learning (IBL). IBL is not a new AI concept, but it failed to scale over the last few decades due to performance issues. Howso uncovered advancements in probability and information theory, as well as accelerated universal search algorithms, which solved IBLs performance issues. This research and development has enabled Howso Enterprise to allow organizations to realize their full AI ROI by truly understanding and taking action against any issues in their AI deployments.
Highlights
- Infer - uncover data attributes and make easy-to-explain inferences to integrate with data catalogs
- Generate - create private, customizable synthetic data and embed watermarks to provide data governance
- Validate - deeply audit data and debug/evaluate existing AI models across your data and AI lifecycle
Details
Pricing
Additional AWS infrastructure costs
Type | Cost |
---|---|
EBS General Purpose SSD (gp3) volumes | $0.08/per GB/month of provisioned storage |
Vendor refund policy
No refunds.
Legal
Vendor terms and conditions
Content disclaimer
Delivery details
64-bit (x86) Amazon Machine Image (AMI)
Amazon Machine Image (AMI)
An AMI is a virtual image that provides the information required to launch an instance. Amazon EC2 (Elastic Compute Cloud) instances are virtual servers on which you can run your applications and workloads, offering varying combinations of CPU, memory, storage, and networking resources. You can launch as many instances from as many different AMIs as you need.
Version release notes
https://support.howso.com/hc/en-us/articles/23336362198295-Howso-Platform-2024-5-0-Release-Notes
Release Summary This release includes a number of performance improvements, most notably improving CPU utilization and throughput for standalone data synthesis and validation using chunk scaling. It also includes official support for Python 3.12 and removes support for Python 3.8. This release fixes installation issues related to TLS and secret management for airgap installations, addresses issues found when upgrading trainee versions, improves privacy when synthesizing data sets that include columns that are hard to predict, and fixes bugs around random number seeds and react calls with case weights.
Synthesizer Improvements Greatly improved CPU utilization and throughput for chunk scaling when running standalone. Strengthened privacy by improved difference evaluation of nominal features values, particularly those that have low predictability, taking advantage of the similarity to increase the entropy of synthesized values and decreasing the amount that the differences of values when considering if the synthetic case is sufficient to be considered a new with regard to the original data. Performance improvements and reduction of memory use. Fixes Removed documentation for previously removed feature attribute time_delta_format. Validator No changes. Validator Enterprise Improvements Performance improvements and reduction of memory use. Howso Engine Breaking Changes Support is officially ended for Python 3.8. Improvements Python 3.12 is officially supported. Improved distance evaluation for nominal features, particularly when the nominal values are not well predictable. Performance improvements and reduction of memory use. Removed smallest fractional p value for targeted analyze, as recent improvements in surprisal space as distances removed the performance benefits brought by the small fractional p value. Improved data ablation efficiency by automatically ablating duplicate cases. Added MCC (Matthews Correlation Coefficient) to prediction statistics. Improved error messages for misconfigured setups. Fixes Fixed rare distance accuracy issue with json, yaml, code, and string distances when comparing against null values. Removed documentation for previously removed feature attribute time_delta_format. Fixed bug where discriminative reacts with case weights would sometimes yield incorrect results. Fixed bug where random seeds would not be changed between runs when loading a new trainee in caml format via the common Python/C-API method. Howso Platform Improvements Upgraded platform components to resolve CVE-2024-1135. Added the trainee worker initContainer resources configuration to the helm chart values. Fixes Fixed issue where the TLS sidecar image was missing from howso-platform kots airgap bundles. Fixed issue where the image pull secret was missing from airgap bundles. Fixed miscellaneous issues with the trainee archive/migration process. Synthesizer Recipes No changes.
Howso Engine Recipes No changes.
Known Issues The Synthesizer UI, currently in beta, currently only supports smaller data sets, and does not expose all of the functionality of Synthesizer. These limitations will be addressed in future releases. When using the Howso Platform and performing reacts with a larger number of features (in the realm of one hundred or more) and requesting significant data from details (e.g., a large number of boundary and similar cases), the scale of the message sizes returned may be notably large and have an adverse effect on performance and memory. This is an edge case that can be worked around by limiting what is requested simultaneously from each react. This is aimed to be fixed in late Q2 or early Q3 of 2024.
Additional details
Usage instructions
- Deploy
- Log In
- Run:
source ~/workplace/howso/venv/bin/activate
- Run:
LC_ALL=en_US.UTF-8
- Run:
LANG=en_US.UTF-8
- Run:
python -m howso.utilities.installation_verification
Support
Vendor support
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.