Take advantage of the many uses of artificial neural networks


The use of artificial neural networks in companies is becoming more and more widespread in the course of the digital development of companies. Despite ethical concerns and challenges in implementing AI, organizations are finding it increasingly difficult to ignore the benefits of using neural networks.

What are neural networks?

Artificial neural networks are a type of data processing that is somewhat based on the structure and processes of the human brain. They consist of an input layer, an output layer and a hidden layer, with backpropagation, an algorithm for supervised machine learning, being a fundamental part of the process.

Organizations mainly employ two types of neural networks: convolutional and recurrent. Each neural network has its own uses.

Use cases for neural convolutional networks

Convolutional Neural Networks (CNNs) are best for solving problems related to spatial data, such as images. Organizations use them for services such as facial recognition software, medical result analysis (x-rays), and image classification on retail websites for targeted marketing.

Ecommerce sites like eBay are using CNNs to create a more efficient buying and selling platform and to improve the customer experience.

“The use of deep learning and visual experiences was an important focus for us,” said Nitzan Mekel-Bobrov, eBay’s chief AI officer.

For example, eBay uses neural networks to automatically list products such as trading cards for sellers based on data from previous listings. Sellers can take a picture of the cards they want to sell and the CNN will identify key characteristics such as card type and issue date and then compile the listing for the seller.

“The vision is really to get to a point where a salesperson can take a picture of anything and we could be able to automatically create the quote for them,” he continued.

With advances in technology, use cases for CNNs increasingly include video components.

For example, CNNs can extract information from video to track data like the number of broken street lights in a city, said Sreekar Krishna, national director of artificial intelligence and director of data technology at KPMG.

The ability to use video analytics to answer these questions can benefit insurance companies. For example, CNNs are well equipped for damage analyzes, crash reconstructions and other forms of spatial analysis.

“You can reconstruct the damage done to either vehicles or even houses … you can actually analyze the effects of the damage by analyzing these videos and images,” said Krishna.

Recurring neural network uses

Meanwhile, the strength of recurrent neural networks (RNNs) lies in their ability to analyze temporal, sequential data such as text and time series data. Common uses of this technology are in speech recognition and prediction.

Retailers and other vendors use RNNs to monitor customer habits and then proactively attempt to retain customers when the RNN detects potential red flags.

For example, RNNs can help identify customers who have made multiple returns in a short amount of time, said Kirk Borne, chief science officer at DataPrime, Inc., an AI-based matching site for applicants and positions.

“You will likely never buy from us again,” said Borne, adding that neural networks can help with the answer, “What can we do to intervene?”

Neural network layer, hidden layer, output layer, input layer
Neural networks consist of an input layer, an output layer and a hidden layer.

Solve business problems with neural networks

When implementing neural networks, “you always need to select a subset of the problem first. And sometimes that can look like a feature … some kind of problem it might solve, ”Mekel-Bobrov said.

Business leaders see the implementation of neural networks as a step-by-step, calculated process.

The iteration is an essential part of defining where and to what extent neural networks should be implemented. For an ecommerce site like eBay, this is true when using AI to create a product listing based on similar listings previously published by sellers. Initially, the data set (i.e., previous listings) can start small so that the neural network can only generate new listings for frequently sold items.

However, as an ecommerce site sells more items and the data set grows, deep learning-based product listing technology can create lists for a greater variety of products and with greater accuracy.

Retailers and financial organizations also often use neural networks to detect fraud.

A neural network can detect out-of-the-norm behavior by noting, for example, when a person who normally buys gas once a week buys it several times a week.

“It’s the combination of the time of day, the product you bought, the amount you bought, the place you bought it – that strange combination says it’s fraudulent now,” said Borne.

The challenges of neural networks

Organizations also need to understand the challenges of creating and using neural networks. Common problems include a lack of visibility into how a model works, bias in the output of a model, and insufficient computational resources to run models on a large scale.

You need to train the people connected to the technology that the technology is not there to make your decisions.

Sreekar KrishnaNational Head of Artificial Intelligence and Head of Data Engineering, KPMG

For large companies like eBay, “the computational resources required are quite challenging … the more advances we make in infrastructure and computing, the faster we can roll out deep learning in many countries.” different use cases, “said Mekel-Bobrov.

A practical example of the ethical problems that arise from relying on deep learning is determining creditworthiness. When a person applying for a credit card is declined by a lender who uses neural networks as the basis for their software, their credit history and financial development are determined by a machine, not a human.

The black box nature of AI becomes an issue when the algorithms’ decisions affect a person’s life, such as whether they are accepted or rejected for a line of credit. A full explanation of why an application passed or failed is required so that employees must understand how a model came to this decision.

However, lenders are increasingly implementing neural networks to analyze potential customers’ financial history and determine creditworthiness. Organizations see significant benefits from implementing and iterating neural networks for recruiting and hiring. The risks associated with neural networks are often insufficient to outweigh the potential improvements in business processes.

How executives can use neural networks

When looking at neural networks, companies should first define the parameters of where neural networks should affect their business and determine where to live and where not, with the consequences of temporarily poor technology performance.

“You have to train the people connected to the technology that the technology isn’t there to make your decisions … Google gave me a bad search result, big deal. My life hasn’t changed. Amazon has the bad one Product recommended, big deal. My life hasn’t changed, “said Krishna.

The most important question IT leaders need to ask their teams is, “Have we measured our value proposition before we get into AI?” said Krishna.

Every industry needs to define its value proposition or goals in order to decide whether implementing artificial neural networks is the right decision.

Krishna noted that “asking which step in the process is causing the greatest pain point in the organization and then using AI to address that problem” is the necessary mindset in implementing neural networks.

However, these inherent risks cannot stand in the way of change. Going AI-first is a common goal in organizations, and those who eschew AI implementation are missing out.

“Some companies just say we’re not going to use AI, but just keep doing things the old way,” said Borne. “But I think these companies are going to lose in the long run because technology advances so quickly that companies that get it right will just win the market.”


Leave A Reply