Accelerated Unsupervised Machine Learning
Nano, the world’s first unsupervised machine learning algorithm designed for high-speed, high-complexity use cases, is enabling new applications in segmentation and anomaly detection.
1,000 faster
Faster than alternative unsupervised machine learning algorithms
High dimensional
Capacity for up to 1000 features in a single model eliminates the need for dimensionality reduction techniques
Auto-tuning
Training parameters automatically scale to optimally fit the data source, so each model can have thousands of clusters if required
No GPU
GPUs are not required to train or deploy models; models train in real time on a single Intel CPU
Designed for highly complex, real-time applications
What is unsupervised ML?
Unsupervised machine learning
In this type of machine learning, the model is trained from unlabeled data, and its objective is to discover patterns, structures, or relationships within the data.
Best use cases:
- Anomaly detection
- Predictive analytics
- Clustering/segmentation
Supervised machine learning
In this type of machine learning, the model is trained from a labeled dataset to create a mapping from input data to output classifications.
Best use cases:
- Classifiers
- Object recognition
- Natural Language Processing (NLP)
Unprecedented performance
The Nano’s speed enables it to train models in seconds that previously would have required hours to days, providing for the first time unsupervised machine learning that can be trained and deployed entirely at the edge.
Superior speed
Do you need to monitor high-speed data in real time or rapidly train and test models to deploy a workable solution? If so, then the Nano is likely a good fit for the use case. The Nano can analyze billions of data points in seconds with consistent inference speed, even for data with a large number of features. Nano was designed for high-speed, real-time applications deployable entirely on-premise or in the cloud.
Built for complexity
If your dataset has hundreds of features and relational complexity, then it’s ideal for the Nano. The Nano can support real-time training and inferencing with models that have up to 5,000 clusters in a 500-dimensional space.
High accuracy
Does your use case require a high level of accuracy? The Nano provides the highest degree of segmentation and anomaly detection accuracy by creating up to thousands of clusters to model the data source. Individualized models can now be trained for each unique data source, providing identification and detection that has the highest specificity and sensitivity.
Flexible deployment
Whether for security concerns or the latency and expense of sending massive amounts of data to the cloud, many customers need to both train and deploy their ML models on-premise or even in an air-gapped environment.
The Nano can be flexibly deployed entirely on-premise or in the cloud. For edge applications, training and monitoring can take place entirely on-premise using commodity Intel SGX-enabled servers.
No GPUs are ever needed.
