There was a time when we thought that we needed more information to solve complex problems. Now, there is so much structured and unstructured data flowing into organizations that it often becomes too complex to reach accurate outcomes in a timely manner. In light of these challenges, Fifth Dimension brings the future to intelligence analysis. Fifth Dimension fuses massive amounts of data from any collection source, in any format, structured and unstructured, to produce automated meaningful intelligence. Using advanced deep learning modules within a mass scale data framework, Fifth Dimension technology actually learns about your data to reveal insights and help you to accurately predict threats, opportunities and trends before they happen.
Fifth Dimension applies deep learning, the most advanced subfield within machine learning and artificial intelligence, based on how the human brain works. Fifth Dimension is the first company with the knowledge and practical experience to apply deep learning to the intelligence domain for analyzing and learning from huge datasets of varied sources in mass. With its deep learning capabilities, the system automatically performs unsupervised learning to uncover hidden patterns and correlations amongst stored structured and unstructured data.
Fifth Dimension’s unique entity model enables learning automatically about millions of new entities, generated by the system, and accordingly turn the ‘unknowns’ into ‘knowns’. The platform constantly accumulates data, analyzes masses of entity activities, cross-correlates insights and identifies new links. The platform’s “brain” performs mapping of entity patterns, anomaly detection and risk ranking. As a result, the system provides the human analyst with prioritized, meaningful intelligence on time.
Data comes in many flavors. There are structured data sources, such as financial transactions, call data records, organizational data sheets and tables, as well as unstructured data sources including images, videos, audio files and documents. Fifth Dimension data adapters are designed to recognize the nature of each data source. They can handle any type of source, any media format and any type of connectivity, be it streaming, network, database, file system, manual upload, or any other. The data from all sources is integrated into a single repository and is prepared for processing.
The data is automatically broken down into small pieces whose structure depends on its data type. Any unstructured data is also handled at this point. Fifth Dimension uses deep learning algorithms to identify hidden elements in the unstructured data – such as faces in images, speakers in audio and entities in text. The Fifth Dimension platform constantly correlates and keeps track of facts and stores everything in one unified model, built on top of the mass scale data infrastructure.
An entity is then automatically created for any object of interest identified in the data, whether it is a person, a vehicle, a place, organization, or any other type defined by the customer’s ontology. Each entity is associated with its entire history of activities, including IDs, images, audio, video, social profiles, charts and documents. This process is applied to known entities, as well as to millions of unknown entities generated automatically by the system. This process repeats continuously as new data is identified and relationships are discovered between entities and their associated data. Through this process the Fifth Dimension platform learns as it expands the dynamic and holistic view of each and every entity in the system and their relationships with one another.
The platform’s ‘brain’ constantly scans all of the data gathered and sorted in the previous stages. The system cross correlates insights to find linked entities, identify similar profiles and pinpoint entities whose evolving patterns reveal insights that no human can find. In addition, using deep learning the system automatically identifies the behavioral patterns of every entity to automatically identify anomalies and breaks in routine. With the risk ranking model the system evaluates the intelligence value of every item and activity in the system, and automatically alerts when an emerging risk detected.
Finally, human analysts interact with the intuitive analytics suite. This gives them access to the insights gained from the data so that they can actively explore, correlate and speculate. Analysts raise their intelligence questions and easily receive precise answers driven by a holistic view of all data sources. Automatic insights received by the system significantly shorten the time to intelligence and maximize analysis production. Together, Fifth Dimension and human analysts can explore the depths of the ocean of data that defines our world and, time after time, emerge holding the treasures that lie below.
Deep learning, the most advanced subfield within machine learning and artificial intelligence (AI), is the closest that computer science has come to creating a learning model based on how the human brain works. Using tens of layers of artificial neurons connected to each other via hundreds of millions of synapses, deep learning has revolutionized AI during the past few years, resulting in 20-30% improvement in all tested benchmarks across multiple domains (vision, speech, text, etc.).
Unlike classical machine learning which relies on manual feature engineering and requires domain experts in each area where it is applied, deep learning completely skips the feature extraction phase, similar to the way our brains process raw data. That is, the best results in computer vision today are obtained by deep learning that operates on raw pixels of images and requires no image processing expertise. The best results in speech recognition are obtained by deep learning that operates on raw audio files without any need for speech recognition experts, and the best language understanding modules rely on deep learning which processes raw characters in any language.
Fifth Dimension is one of the leading companies in the world employing deep learning at the core of all its modules. The deep learning research group within the company comprises more than a dozen experts in the field of deep learning (all holding PhD and MSc degrees), some of whom are renowned researchers in the field of deep learning with dozens of publications in leading AI journals and conferences.
Using its advanced deep learning technology, Fifth Dimension is the first company that is capable of analyzing and learning from huge datasets of varied sources (mass data). Thus, highly complex non-linear patterns across the whole mass scale data platform are efficiently uncovered by Fifth Dimension, which results in real-time actionable intelligence.
Deep learning, being completely agnostic to the input, is capable of efficiently learning and processing problems across any domain. Specifically, the results obtained by deep learning in computer vision, speech recognition, and natural language understanding surpass those obtained by other methods by a wide margin (20-30% improvement).
Relying on its deep learning capabilities, Fifth Dimension is capable of fully processing and learning from unstructured data (images, videos, audio, text, etc.), thus, extracting meaningful insights and knowledge from those areas.
In most data-intensive organizations, the vast majority of the data is unstructured, and is not processed and analyzed (typically only structured data, in the form of databases is analyzed). Fifth Dimension provides human level performance for face recognition, speaker recognition, and multiple tasks for natural language processing (text similarity, recommendation, translation, sentiment analysis, etc.). This learning accuracy, coupled with a processing speed which is many millions of times faster than that of humans, allows Fifth Dimension to successfully incorporate any unstructured data within the body of knowledge that it automatically forms.
One can imagine Fifth Dimension’s mass data platform as a huge library, where all pieces of knowledge are orderly stored on the shelves. Any knowledge extracted from the unstructured data (due to deep learning) is also placed on the same shelves alongside the structured data. However, this huge library is not inanimate. While Fifth Dimension provides the most sophisticated tools for manual interaction and analysis, behind the scenes it is constantly performing unsupervised learning to uncover additional patterns and correlations amongst the stored data.
Deep learning has been proven by far to be the best approach for unsupervised learning, especially since the vast majority of interesting patterns are nonlinear, and thus nearly impossible for humans or other linear models to uncover. Running its unsupervised deep learning module around the clock, the huge library of Fifth Dimension constantly grows even without any human interaction. Whenever the system uncovers novel patterns that it deems relevant for the human analyst, only then a corresponding alert is provided together with the full set of supporting data and the related connections.
Everything we do creates an avalanche of data. But as the wealth of data grows, so does the range of tools that we use to try to make sense of the data. This leads to two types of challenges. First, maintaining an efficient storage solution for all that data becomes an ongoing ordeal. Second, effectively analyzing so much data using traditional technologies can take hours, days or even weeks to provide answers. By then, it is often too late to act.
There is already a wide range of tools for collecting everything – IDs, records, posts, logs, databases, biometrics, etc. But the main intelligence gap is caused by the lack of tools to enable taking full advantage of all the data that is generated. Fifth Dimension resolves data challenges which, because of their sheer size, complexity and multiple sources and formats were previously impossible to overcome.
Fifth Dimension fuses the massive amount of information collected from a variety of sources, structured and unstructured, into a single unified model, supported by an intelligent Data framework. Fifth Dimension preprocesses data in a way that makes the information exponentially easier to search and analyze compared with any traditional method for both machines and for humans. The system converts masses of stored multi-sourced data into usable and accessible actionable insights – continuously and automatically.
Fifth Dimension provides a distributed mass data computing-based framework that eliminates the bottlenecks of traditional databases. Using commodity hardware, Fifth Dimension technology enables ultra-fast access and processing of PETABYTES of secure on-site data to generate insights that you need in order to address time-critical problems.
The Fifth Dimension Data Framework is based on cutting-edge Apache Spark distributed computing database technology, integrated into a Hadoop cluster. Together, these technologies enable easy storage and processing of huge amounts of data in parallel. The Fifth Dimension platform is designed for quick deployment process and allows maximal customer independence, with easy maintenance procedures.