The Ubex Technology Base

May 21, 2018

The application of advanced neural and Artificial Intelligence technologies on a broader basis in various markets has long been considered a primitive luxury as the mastodons of the AI industry have always been content with a low level of productivity of their constructs. By monopolizing the market of AI and producing low yield constructs for high fees, the developers were able to slow the rise of hi-tech startups for a very long time.

The situation started changing with the advent of blockchain technologies as independent developers, such as Ubex, were able to gain access to massive datasets and decentralized solutions for relatively low costs, thus allowing for the widespread development of independent, hi-tech solutions for businesses. By making the transition to blockchain and effectuating the merger of AI and neural networks, Ubex was able to bring an entirely new dimension of transparent, efficient and large scale solutions to the advertising industry, thus benefiting publishers and advertisers alike.

The Ubex system core (Advanced Decentralized Advertising Marketing – ADAM) is responsible for making the decision to display advertising materials on the blockchain platform, thus eliminating the human factor from the equation and completely automating the process of advertiser and publisher matching, effectively placing ads for maximum audience targeting.

For advertisers, the most common feature available on the Ubex platform is the widget, which is visible to all participants of the advertising process, including the publisher and the end user, or the viewer of the ad. The widget appears in the system after the advertiser registers and adds an offer that advertises the services or products they wish their target audience to see. Along with other promotional materials, the advertiser registers a widget, loads the start banner and specifies the format and list of fields requested from the visitor. In addition, the advertiser specifies the parameters for displaying the widget, such as the subject, geodata, temporary data, a list of visitor interests, age and censorship restrictions, etc. Given the fact that the system is decentralized, the advertiser can register at any partner sites that are members of the Ubex exchange network. The data of the advertiser, the offer and promotional materials along with the parameters for display are added to the blockchain system and can be used by any member of the exchange, provided the necessary rating level allows it.

The Ubex system core (Advanced Decentralized Advertising Marketing — ADAM) is responsible for making the decision to display advertising materials on the blockchain platform

The widget can start working on the publisher’s site as soon as the publisher registers, passes the KYC procedure on the Ubex platform and adds a placement resource. When creating a site, the publisher, similar to the advertiser, specifies a set of parameters for the widget. To achieve better results, the publisher can integrate the tracking services of the exchange, thus providing extended statistics to authorized users. Decentralization, as in the case of the advertiser, allows one to register at any of the sites of partners available on the Ubex network. Site data, its parameters and integration settings with tracking services are included in the blockchain system and can be used by any decision-making service that operates within the Ubex exchange with the help of an advanced AI construct.

Upon completion of placement of the widget, an initial request for the display of advertising material is sent to one of the hundreds of tracking services of the Ubex advertising exchange. The exchange will operate dozens of automated decision-making systems based on neural networks, or DSPs — Demand Side Platform. The first such system starts working after the visitor’s direct visit to the platform website. The tracker marks the request for further links and collects all available data that came with the request, passing it for analysis to a free DSP.

The DSP receives dozens of parameters at the input stage, including website data, current user data, extended data on authorized users in case of integration, and the history of advertising displays for the specific user. The tracker marks current users in the browser, understanding when the same user requests displays of advertising material even on another website. In addition, the history of the user’s orders on the publisher’s website is also provided, for example, in the case of online stores, this includes previous orders of the user and other information available relevant to that particular user. The more input data is provided, the more accurately the neural network will be able to understand the interests of the user. At the output stage of the DSP, the result is a detailed map of user preferences cluttered on an accessible list of topics of the exchange.

The next DSP, or DSP-2, selects the list of potential candidates, such as advertising materials for display, whose advertising can be potentially interesting to the user. By receiving input from the previous DSP, the current DSP-2 conducts its regression evaluation of all available advertising materials and outputs an estimate for each. As a result list of possible promotional materials is returned to the tracker that leads the current query.

DSPs will perform regressive estimation of qualitative or quantitative indicators for compliance with a certain result

Most existing programmatic exchanges would have already decided to display what they would have deemed as the most relevant query. But not Ubex.

At the input stage, the next DSP-3 receives all the parameters collected by the tracker on the current query and selects one and only one result from the selected materials. It selects not only the most relevant, but also the most financially effective result, which is beneficial both for publishers and advisers. For this purpose, neural network regression is performed for each user action on the advertising material and a map is created containing the estimated indicators.

At the next stage, the tracker already knows which banner or widget to display and passes the task to one of the thousands of free renderers to build a script / css / images to provide a final reply to the user. The final bundle of js code is returned to the tracker, which in turn returns the result as ready-made advertising material to the visitor of the publisher’s site.

All further user actions, including analytics, clicks and actions with advertising material are further served by the tracker leading it and the AI and neural network base. Thus, we obtain a highly organized, intelligent system consisting of hundreds and thousands of DSPs that make small decisions at every stage of displaying advertising material.

As part of the system, thousands of DSPs will perform regressive estimation of qualitative or quantitative indicators for compliance with a certain result. Each DSP is a neural network. The layer of input parameters consists of normalized per unit indicators. At the output stage, the neural network gives normalized per unit judgments about the set of output properties.

Each node within the hidden layers of the network is a neuron. At the first approximation, the artificial neuron imitates the properties of a biological neuron. At the input stage, an artificial neuron receives a certain number of signals, each of which is the output of another neuron. Each input is multiplied by the corresponding weight, similar to the synaptic force and all products are summed, determining the activation level of the neuron. The input signals correspond to signals that come to the synapses of a biological neuron. Each signal is multiplied by the corresponding weight w1, w2, …, wn, and is sent to the summation block, denoted by Σ. Each weight corresponds to the “strength” of one biological synaptic connection. The summing unit, which corresponds to the body of the biological element, algebraically combines the weighted inputs, creating the output NET.

Thus, due to the use of neural networks, the Ubex algorithm allows for taking into account a variety of factors that affect both the relevance of the advertising display and its economic efficiency. The Ubex algorithm collects data about the profile of each specific user, including their behavior, geography, visiting time, interests, etc., and calculates the probability of a targeted action for each individual advertisement. The more data passes through the Ubex neural networks, the more effectively they solve their tasks.

By combining the most advanced technologies in in neuromation, Ubex is creating a unique platform aimed at maximizing profits, minimizing waste and turning the costly process of buying advertising into an immediate decision of selecting target audiences, benefiting all participants of the advertising cycle.

Comments