Research & Development
AI data analytics systems
By time series in artificial intelligence we mean digitised sequences. The problems encountered when processing time series can be grouped into three main categories:
(1) Sequence-to-sequence (seq2seq) training, (2) time series forecasting, and (3) time series labeling and segmentation. In general, the AI models we use are applicable to each of the aforementioned problems, but they are not equally good at solving them, each AI model has a problem type that it is designed to solve best.
Basic problems of AI data analysis
AI-based time series analysis
The problems encountered when processing time series can be grouped into the following main categories:
(1) Sequence-to-sequence (seq2seq) training, i.e. the prediction of time series that are assumed to be related to each other:
These cases are characterized by the fact that both the input and the output are some kind of sequential (time series) data.
(2) Time series prediction, i.e. prediction of unknown points in a time series:
In general, this is a future point, but it may also be a prediction of a value associated with an intermediate missing point in time (e.g., an error or hidden information that is dropped due to an external factor).
(3) Time series labeling and segmentation, i.e. assigning descriptive values to the entire time series or to certain points in it:
Such a task may be, for example, to predict some malfunction or to detect certain patterns within a sequence.
Sequence-to-sequence (seq2seq) training
Time series prediction
Time series labeling and segmentation
TIME SERIES FORECASTING AND BUSINESS INTELLIGENCE (BI)
Forecasting time series data – LSTM, GRU, SNN
According to the current state of science, the most suitable solutions for time-series analysis and time-series forecasting are the LSTM – Long-Short Term Memory and the GRU – Gated Recurrent Unit recurrent neural algorithms. In addition to these, an emerging technique is Deep Learning’s Attention-based algorithms, which excels in time series transformation. These algorithms are the so called Transformer algorithms. In cases where a method needs to be applied to data from diverse domains, the use of DNC – Differentiable Neural Computer memory-augmented algorithms may be recommended, and so it is proposed to examine their use on time series from various types of enterprise data.
In addition to the previously mentioned models, there is another promising way towards a more efficient (“Green”) AI technology. This is the use of dynamic neuromorphic neuron models running on specialised neuromorphic chips (chips mimicking the operating principles of the human nervous system), which can reduce the cost of running the AI models. Such neuron models are implemented by SNN – Spiking Neural Network algorithms.
Our researchers hypothesize that a combination of the aforementioned algorithms may provide the most promising results for the design of AI-based business intelligence (BI) systems, since LSTM is generally more effective in a long-term context, while GRU is usually more efficient in short-term contexts. Their combination therefore is assumed to be suitable for drawing correct conclusions in both short and long term scenarios. Successful implementation of the Transformer algorithms may further improve the results, while a memory-augmented neural algorithm may be able to function as a generalised model that can be used in a more generic way.
Most efficient forecasting algorithms: LSTM, GRU
Malfunction and anomaly detection capability
Optionally using the latest neuromorphic chips and SNN algorithms
State-of-the-art business intelligence (BI) systems
Research and Development with Trilobita
Phases of our R&D projects
R&D Project planning
In the R&D project planning phase, we help our clients find the most optimal use of resources. We prepare the financial and technical design of the project and prepare the proposal for the selected funding scheme.
Applied research
In the applied research phase, we prepare the necessary research plans. We carry out and document the series of experiments based on our research methodology.
Evaluation of research results
We evaluate the results of the series of experiments using various data analysis methods and prepare the research summary document.
System Planning
Based on the research results, we design the systems for our customers. We use our own system design methodology and tools for the planning.
Development and testing
Our development methodology combines elements of classic waterfall and agile methodologies, flexibly adapting to the needs of the given client and project. The efficiency of our development and testing work is further enhanced by a number of our already tested, ready-to-use system modules.
Support
After the completion of our R&D projects, we always provide follow-up and support services to our customers for the solutions we have delivered. Our goal is to establish successful, long-term partnership with our clients.
We believe that every hour spent on design pays off many times over in the implementation and roll-out of our systems. Our ergonomically designed user interfaces provide our customers with a new user experience and ease of use.