FAQ
See the section below for answers to frequently asked questions. If you require any further information, feel free to contact our support team at any time.
COR-RiSTOR SaaS
What types of data can be analyzed with your software?
You can use any data that is suitable for solving a classification task. The columns for the input data can contain floating point or integer numbers, as well as dates. The column that describes the class to be recognized can also contain character strings. Application examples: weather monitoring, system monitoring, fitness monitoring.
How can I prepare my data for clustering?
You can clean, scale, and transform your data to prepare it for clustering.
What are the advantages of correlation analysis of my data?
Clustering helps identify patterns and structures in your data, understand correlations between data, and support decision-making processes.
How do I interpret the results of the correlation analysis?
The pairwise correlation between two given input columns is represented as a correlation function that depends on both input columns. As a result of the analysis, for each combination of input columns there is a set of polynomial parameters for the correlation function. The polynomial parameters describe the correlation function and are used to link the two given input columns according to the correlation function. This means that two input columns are transformed into an output column using the correlation function before the actual clustering and classification is carried out.
Are there limitations regarding the amount of data?
For an initial analysis of your data, we need to impose a few limitations to ensure the analysis process does not take too long. Therefore, your data can contain a maximum of 4 columns for the input data and 1 column for the class information. The number of rows per column is limited to 10,000. The number of classes is limited to 10 classes.
Can I export or visualize the results of clustering?
Yes, you can export the results and save them in various formats (PDF, JSON). We also offer visualizations that graphically represent the correlation between two given input columns.
Does your software provide support or guidance for beginners?
Yes, we offer extensive documentation and support to help you get started with clustering of pairwise correlated data strands.
COR-RiSTOR
What is the COR-RiSTOR?
With COR-RiSTOR you can preprocess all your sensor data in real time on the edge – with a strongly increased cluster recognition rate!
The COR-RiSTOR SaaS enables the identification of linear and non-linear correlations of all possible pairs of sensor data directly with a strongly increased cluster recognition rate and a fundamental minimization of hardware and energy requirements, massively reducing the need for cloud computation of this data. The extracted correlation function of a given pair of sensor data is then configured into the TiF-MEMRiSTORs in the COR-RiSTOR.
The COR-RiSTOR contains the same number of TiF-MEMRiSTORs as correlation between pairs of sensor data have to be computed in real time. The TiF-MEMRiSTORs are specifically designed and operated for high-precision pairwise correlation analysis of your sensor data.
What does edge sensor data mean in this context?
Edge sensor data is defined as sensor data preprocessed at the sensor.
How does the COR-RiSTOR work?
The quality of measurement and sensor data can significantly be improved by clustering the sensor data with the COR-RiSTOR® algorithm. First, using COR-RiSTOR SaaS the linear and non-linear correlation between a given pair of sensor data is extracted. Then the TiF-MEMRiSTOR in the COR-RiSTOR who will compute this correlation between the data of the given pair of sensors will be configured correspondingly with support of our team.
We provide recommendation how to configure and operate the TiF MEMRiSTORs in that way that they compute significant linear and non-linear correlations of pairs of your sensor data in real time. Subsequently, the user will be able to perform pre-processing and correlation analysis of sensor data in real time and cluster the sensor data with increased cluster recognition rate in comparison to the clustering of unpreprocessed sensor data.
What is the difference between the COR-RiSTOR and other systems for correlation analysis?
The COR-RiSTOR is an analog electronic device for correlation analysis of data in hardware and in real time between data from each pair of sensors in a transaprent manner. In addition to resource-efficiency and transparency also monitoring of sensor data is possible, i.e. with COR-RiSTOR you can also detect unusual and drifting sensor data, thus enabling fully automatic sensor data processing for the first time.
How long will the correlation of COR-RiSTOR work?
After having configured the TiF-MEMRiSTORs of the COR-RiSTOR, the TiF-MEMRiSTORs compute pairwise correlation of sensor data in real time according to their configuration. If you want to compute another correlation of sensor data with a given TiF-MEMRiSTOR, you have to reconfigure this TiF-MEMRiSTOR so that the other correlation is computed.
Why do I need this as a developer, why should I be interested and why should I order the COR-RiSTOR?
We expect developers who wish to analyze and cluster very dynamic input sensor data in a resource-efficient, transparent manner and who on top are looking for input data monitoring will be among first test users.
What is the COR-RiSTOR able to analyze, do you have test projects?
Yes, we are working on pilot projects in the area of monitoring production systems, e.g. conveyor belts, where we first determined correlations using COR-RiSTOR SaaS and now compute and monitor correlations for vibration, weight, and belt speed. Another example are railway bridges, where we first determined correlations using CORRiSTOR SaaS and now compute and monitor correlations for temperature and tension.
I would like to know if the device can be trained with new correlation functions?
If there are new data sets with different significant correlations, we recommend to determine changes in correlations first using COR-RiSTOR SaaS and then reconfigure corresponding TiF-MEMRiSTORs of the COR-RiSTOR. Customers can expect our support here, e.g. with the COR-RiSTOR SaaS or with the reconfiguration of the TiF-MEMRiSTOR.
How does the data get in real time into your demo board?
The board has interconnects to connect the sensors to the COR-RiSTOR demo board in real time.
Why do I need that hardware if I can run the code on my existing devices?
The COR-RiSTOR demonstrator stands out as a unique electronic product designed specifically for real-time pairwise correlation analysis of sensor data at the edge using the innovative TiF-MEMRiSTORs. What sets it apart, is the unparalleled configurability, robustness against sensor data noise, and real-time, unclocked computation of the TiF-MEMRiSTORs, offering a level of flexibility not found in conventional software-based solutions. This distinctive hardware approach enables more efficient and precise correlation analysis. This distinctive feature ensures a significant reduction in the overall energy consumption during data processing. In essence, the COR-RiSTOR introduces a innovative hardware solution that redefines the landscape of correlation analysis in edge sensorics, offering advantages in configurability, real-time operation, and energy efficiency that are unparalleled in the current market.
Is it a kit one builds?
No, it is not a kit one has to build. The COR-RiSTOR board is an electronic circuit in a housing and with interfaces for input and output data. The COR-RiSTOR board contains our core innovation – the TiF-MEMRiSTORs for pairwise computation of linear and nonlinear correlation between sensor data. The TiF-MEMRiSTOR is a passive electronic component which stores and process data in the same cell without data transfer.
In case I would like to get specific information about the COR-RiSTOR with the mobile phone, is there a special program needed?
At the beginning our I/O software will be available just for desktop. It can be used to visualize COR-RiSTOR results. We plan to further develop our I/O software system independently.
TiF-MEMRiSTOR
Is the TIF-MEMRiSTOR just another memristor?
The TiF-MEMRiSTOR is a non-linear memristor which can process and store (analog) data in the same cell without data transfer. TiF-MEMRiSTORs in the CORRiSTOR compute the correlation between pairs of input data. Other, linear memristors can store, but not process, data.
What does "analog hardware" mean in this context? Does the device use analog input signals? Or is the internal computation based on analog circuits?
The TiF-MEMRiSTOR is an analog stand-alone device which computes analog output from analog input signals.
Can you tell me more about the previous research and are the publications about your technology in per review papers?
If you are interested, you can take a look at a reference on TiF-MEMRISTORs in comparison to other memristors [https://pubs.aip.org/aip/jap/article/135/20/200902/3295370/Prospects-for-memristors-with-hysteretic] and at a reference on the physical model describing how to store and how to process data in the TiF-MEMRiSTOR [https://arxiv.org/abs/2402.10358v2].
Where can I get customized TiF-MEMRiSTORs?
Customized TiF-MEMRiSTORs are available as stand-alone TiF-MEMRiSTORs. We recommend to test stand-alone TiF-MEMRiSTORs together with the MeasureKit. Please contact us, if you plan to run a customer specific COR-RiSTOR-based product development project.
Will the TiF-MEMRiSTOR have a usage / shelf life like other microchips?
Our chips will have their usage / shelf life time. We are currently working on determining these parameters.
Do degraded chips mean having to buy a new device?
It is a plug-in TiF-MEMRiSTOR chip carrier which can be easily exchanged if a TiF-MEMRiSTOR chip needs replacement.
Can we purchase just the chips? Do you have schematics and examples?
If you are interested in the stand-alone TiF-MEMRiSTOR chip, please contact us directly.
How will you distribute your product? Will it be possible to buy chips directly from you, or from one of the major electronic components suppliers?
Yes, we also plan to directly provide TiF-MEMRiSTORs. If you are interesed in bying TiF-MEMRiSTORs directly, please contact us so that we can send an offer on the TiF-MEMRiSTOR chip carriers.
How do the input signals get in real time into the device? I don't see any input connectors.
The memristor is a two-terminal-device. Both terminals are input and/or output connectors.
Isn’t this adding to the energy consumption since we still need a computer to use this?
The performance of the TiF-MEMRiSTOR device as the core element of the COR-RiSTOR demonstrator is unique. The reduction in energy consumption becomes more visible the more dynamic the data to be clustered is. E.g. for multiplication, it comes with a 90% reduction in energy consumption.
