SODA Ltd is an expert on classification tasks where data is scarce.
SODA implements Transfer Learning techniques.

UNetNeueral Networks for medical and agricultural implementations. UNets are an
excellent segmentation tool providing a good training set. When samples are scarce
it is a work of art to create artificial samples. This is one of the expertise of SODA
Ltd.

Python GUI includes environments such as TKINTER and Kivy.

An adversarial generative network (GAN) is a class of machine learning frameworks designed by Ian Goodfellow and his colleagues in June 2014. Using a training set, this technique generates new data with the same statistics as the training set. Using GANs trained on photographs, for instance, it is possible to generate new photographs that appear at least superficially authentic to humans. While GANs were originally developed as a generative model for unsupervised learning, they have also proven useful for semi-supervised learning, fully-supervised learning, and reinforcement learning.

SODA Ltd, performs complex regression tasks while using Gradient
Boosting Trees, Random Forests, BART, LSTM and ordinary MLP Neural Networks. In
addition, SODA implements proprietary training methods.

Unsupervised learning includes High Order Clustering and
Gaussian Mixture Models GMM. The input for these unsupervised methods usually
comes from preprocessing algorithms that include artificial neutral networks. In that
case the input is also known as Embedding

PyBind11 allows to add new functions to
python in a straight forward way. The memory that is allocated by the C++ code is
freed by the Python because the returned objects are Python objects. PyBind11 is a
powerful tool that allows fast C++ code to run hundreds of times faster than in
Python alone. For recursive tasks, an average of 380 times faster code than in
python is a typical performance of a C++ code.

A form of outsourcing called data science as a service (DSaaS) involves delivering information from advanced analytics applications, which are run by data scientists at an outside company, to corporate clients for use in their business processes. Client data is collected, prepared for analysis, and then algorithms are applied to the refined data. The results of the algorithms are returned to the customer. Customers typically upload their data to a cloud database or a big data platform, where a team of data engineers and data scientists can analyze it.

SODA Ltd performs time series regression including but not
limited to LSTM, GRU, Arimax, SARimax, exponential filtering, and ordinary MLPs.
SODA Ltd performs flexible Dynamic Time warping with Dynamic Programming
Siamese Networks where Siamese Networks matching is not brittle.

Decision trees include Gradient Boosting Trees, random Forests and
Bayesian Adaptive Regression Trees. The same tree structures can be used also for
regression tasks. Decisions trees are usually more versatile than regression trees
because unlike in regression trees the output values are bound.

Optimization includes stochastic methods such as proprietary Genetic
Algorithms that are designed to avoid evolutionary pits by a supervisor algorithm.
Optimization includes methods such as Particle Swarm Optimization and local
minimum action.

Every business today relies on data. Information about the business is used to answer questions such as: What’s a new customer worth? What are the fastest-growing product lines? Medical research submission to Helsinki and more. Since the data is stored in different formats and managed by different technologies, data analysis is challenging. In spite of this, the tools used for analysis assume that the data will be managed by the same technology, and stored in the same manner. To answer critical business questions, businesses of all sizes have to comb through enormous amounts of disparate data. The goal of data engineering is to support the data-consuming process, enabling analysts, data scientists and executives to securely, quickly, and reliably inspect all available data. Data engineering helps make data more useful and accessible to consumers. Data from each system needs to be sourced, transformed, and analyzed by data engineering. A relational database, for example, manages data in tables, like an Excel spreadsheet. Tables contain rows, and rows have the same columns. Depending on the data being stored, a customer order may span dozens of tables. This is why even simple business questions can require complex solutions. Understanding the data and technology behind each system is essential. After data engineering sources and curates the data for a given job, it is much easier for consumers of the data to use.