Data Filtering

YOUR PRE-TRAINING DATA CLEANING ASSISTANT 

 

You let the SDK orchestrate training experiments

You set up the Alectio SDK on your system and let the SDK orchestrate a series of short training experiments. The system generates a data filter (a small predictive model capable of scoring data as per its value in the context of your model). Data filters are also relatively model-agnostic so you can reuse the filter with other variations of your model.

Use your data filter to get new training data

You download the data filter on your system and use it to filter new training data prior to (re)training your model with it. You can also deploy the filter on the edge of an IoT device to decide in real-time whether a given record needs to be stored/transferred, or if you can delete it.

A smaller and optimized subset of data that is just right for your model

You can now pre-process any dataset to reduce its size before training, which enables you to train the same model much faster and much more efficiently, but also to flush any useless data out of your storage instances. The filtering process is tuned to your use case and model and ensures that no bias will be induced during the process. Data filters can be used for data management, improved data collection, data cataloging and more!

At the edge of your model sits how your training data is working or not working for your models. Alectio’s Data Filtering can in real-time listen and learn from your models what data they need to perform at their expected best. This can be especially useful for use cases like autonomous vehicles, where large volumes of data are collected and stored in the cloud and where labeling costs are especially high.

 

Let Your Models Tell You What Data it Needs!

Alectio’s Reactive Data Filtering gives you real-time feedback on the most useful data for your models.

Works with your existing data sets to predict which data is most useful for your models.

Works to determine model performance with a given training data set – even without internet access.

Works without needing to collect every byte of data. We’ll help you save only the most important information–in real-time.