Machine learning


Today, AI is everywhere. From detecting all those pictures of cats to making recommendations for online shopping, people are surrounded by AI-powered products and services.

The last decade has been an exciting time for Artificial Intelligence, as it has delivered impressive results in areas ranging from image detection to self-driving cars.

Most people are familiar with AI performing tasks on our behalf like powering online customer support as seen on Zendesk, Amazon Alexa or even recommending products as we do so on e-commerce platforms like Shopify. There is a lot of excitement around what is coming next and how AI will continue to impact our lives in the future.

Meanwhile, deep learning—a powerful form of machine learning that's used to train neural networks and improve their performance—has helped these systems deliver increasingly accurate results. However, while they might be useful at this point in time, AI and deep learning aren't exactly easy to understand or use for the regular person; they're not very user friendly.

An important issue in regard to AI models is their explainability.

When it comes to machine learning, we're only starting to realize that this is a problem. We can't understand why algorithms do what they do—the black boxes of AI are making decisions for us, and we have no idea what makes them tick. Well, that's not entirely true.

The data that goes into the algorithm is being evaluated and weighed by the algorithm in order to make its decision. But the way the algorithm does this is still a mystery—even to the people who wrote the specific machine-learning algorithm in question.

We're facing an increasing lack of accountability with most of AI today, and regulators are taking note.

“Right to Explanation” is one of the key elements of the GDPR of European Union.

There are also many other acts from other countries and jurisdictions which include the requirements for explanability.

These acts are Equal Credit Opportunity Act (ECOA) and the Fair Credit Reporting Act (FCRA). Privacy-preserving machine learning

Another important part that concerns Machine Learning is privacy. Machine learning (ML) has influence on data privacy in several important ways.

First, when we train the machine learning models, we can be using data in the training data set that contain sensitive personal information. Also, because the machine learning models are the better the more data we have, this problem only increases with the main goal of ML - get more data.

The other way how machine learning models are involved in privacy is because the ML models make predictions or suggestions about individual users. So they are part of the process. This means there is high potential for data privacy violations when the machine learning models are deployed in production.

Data privacy violations happen in different way :

- personal data from data sets can somehow migrate to the weights of the models

- it is possible to extract sensitive personal data from the model by repeatedly using the machine learning model

- groups of persons, e.g. outliers, can be more easily identified than others

- with repeated usage of ML model, one can learn [a] person’s image from name (what is known as so-called Model inversion) or [a persons] presence can be detected in training data (Membership inference)

- ML outputs can be combined with side information (e.g. from public records) to reconstruct personal information

For this reason, methods have been developed to address the problem of data privacy in artificial intelligence field. One of these ideas is differential privacy, favoured by companies such as Apple and Google.

Explainable Artificial Intelligence Software

For data sciencists, explainable machine learning (XAI) is implemented in many open source and commercial software packages.

Well-known packages for XAI practitioners include:

  • SHAP
  • LIME
  • ELI5
  • Skater
  • PDPBox
  • InterpretML

Machine learning

Machine learning (ML) is one of the subsets of artificial intelligence. The goal of supervised ML is to help machines accomplish tasks through their own experience. Instead of coding specific instructions on how the machine should accomplish a task, we give it examples to learn from and improve with experience. The ML algorithm learns by ingesting data and looking for patterns in that data.

An example of ML model is one that is able to do a sentiment analysis of tweets to produce fear and greed index for cryptocurrencies, stocks or other financial securities.

At its core, machine learning gives computers the ability to learn without being explicitly programmed. In other words, we don't have to tell the computer how to perform certain tasks: we just let it figure out how to do them on its own. So instead of coding specific instructions for the computer, we give it various examples, and then tell it which ones are "correct" or "incorrect."

The machine will then use those examples to complete a task autonomously. The whole idea behind machine learning is that you can set up a system so that computers can learn from their environment without any explicit programming. This might seem counterintuitive at first, but keep in mind that this is exactly how humans learn. We don't have to consciously think about all the steps involved in making breakfast in order for us to make breakfast; we just do it because we know how based on what we've learned before.

Machine learning can be a great tool for finding best keywords of niches, by relying on factor approach to keywords research, e.g. by taking into account SERPs, search volume, on-page optimization and other factors.

Product Categorization

Another important machine learning task is product categorization in ecommerce domain.

Did you know that product classification is a well-studied discipline within the field of Natural Language Processing? Most people don't—but it's true! Product taxonomy is a complex process, and it may sound simple: put shoes in the 'footwear' category and pliers in 'tools', and so on. Still, the actual reality is far more complicated.

We will discuss the importance of product categorization, the challenges it brings to e-commerce businesses and how machine learning (ML) can help such businesses.

Product taxonomy is an essential component of a successful online shop. Without it, consumers have no way of navigating through a website and finding the products they want. This makes product categorization critical for e-commerce businesses: if your product categorization is poor, you will lose customers and sales. But how do you make sure your product placement is done right? This is where machine learning comes in.

Product tagging

Another important way to improve the sales on your ecommerce shop by using machine learning models is product tagging.

Unlike product categorization which assigns one label per category tier, ecommerce product tagging assigns more than tags to given product.

Here are a few examples for product names and their generated tags:

geo quartz earrings →earrings, quartz, rose quartz, stone, gemstone, jewelry, sterling silver, crystals, clear quartz, crystal, earring, blue, gemstones, gold, natural stone, silver, crystal jewelry, amethyst, crown chakra, pendant

candied pecans → nuts, almonds, gluten free, vegan, chocolate, baking, dark chocolate, candy, fruit, caramel, dessert, snack, snacks, sweets, roasted, gourmet, local, milk chocolate, white chocolate, food


Interesting links on variety of topics

Cryptocurrency heat map: crypto heat map app

Bitcoin: Bitcoin

Cryptocurrency API for social media sentiment, technical analysis, news and other topics: crypto sentiment analysis api

Useful crypto coin comparisons


Bitcoin is a digital currency that was invented by an unknown programmer, or group of programmers, under the name Satoshi Nakamoto. It was released as open-source software in 2009.

Bitcoins are created as a reward for payment processing work in which users offer their computing power to verify and record payments into the public ledger. Called mining, individuals or companies engage in this activity in exchange for transaction fees and newly created bitcoins. Besides mining, bitcoins can be obtained in exchange for fiat money, products, and services. Users can send and receive bitcoins electronically for an optional transaction fee using wallet software on a personal computer, mobile device, or a web application.

What is Ethereum?

Ethereum is a platform and a programming language that makes it possible for any developer to build and publish next-generation decentralized applications. Ethereum can be used to codify, decentralize, secure and trade just about anything: NFTs, domain names, financial exchanges.

Hi there! That's not a question we encounter every day—what is Ethereum?

Well, first and foremost, Ethereum is an open-source technology that runs smart contracts. What are smart contracts, you ask? Well, those are little computer programs that live on the blockchain and run exactly as programmed without any possibility of e.g. downtime.

We're talking about unbreakable self-executing deals that can move around value like money—no middleman needed. Cool stuff, right? Well, maybe not to everyone—Ethereum also has a massive community that spans the globe. And it's growing every day.

Ethereum vs Bitcoin Ripple vs Bitcoin Binance coin vs Bitcoin Solana vs Bitcoin Cardano vs Bitcoin Polkadot vs Bitcoin Dogecoin vs Bitcoin
Bitcoin vs Ethereum Ripple vs Ethereum Binance coin vs Ethereum Solana vs Ethereum Cardano vs Ethereum Polkadot vs Ethereum Dogecoin vs Ethereum
Bitcoin vs Binance Coin Ethereum vs Binance Coin Ripple vs Binance Coin Solana vs Binance Coin Cardano vs Binance Coin Polkadot vs Binance Coin Dogecoin vs Binance Coin
Bitcoin vs Tether Ethereum vs Tether Ripple vs Tether Binance coin vs Tether Solana vs Tether Cardano vs Tether Polkadot vs Tether Dogecoin vs Tether
Bitcoin vs Solana Ethereum vs Solana Ripple vs Solana Binance coin vs Solana Cardano vs Solana Polkadot vs Solana Dogecoin vs Solana
Bitcoin vs USD Coin Ethereum vs USD Coin Ripple vs USD Coin Binance coin vs USD Coin Solana vs USD Coin Cardano vs USD Coin Polkadot vs USD Coin Dogecoin vs USD Coin
Bitcoin vs Cardano Ethereum vs Cardano Ripple vs Cardano Binance coin vs Cardano Solana vs Cardano Polkadot vs Cardano Dogecoin vs Cardano
Bitcoin vs Ripple Ethereum vs Ripple Binance coin vs Ripple Solana vs Ripple Cardano vs Ripple Polkadot vs Ripple Dogecoin vs Ripple
Bitcoin vs Polkadot Ethereum vs Polkadot Ripple vs Polkadot Binance coin vs Polkadot Solana vs Polkadot Cardano vs Polkadot Dogecoin vs Polkadot
Bitcoin vs Terra Ethereum vs Terra Ripple vs Terra Binance coin vs Terra Solana vs Terra Cardano vs Terra Polkadot vs Terra Dogecoin vs Terra
Bitcoin vs Dogecoin Ethereum vs Dogecoin Ripple vs Dogecoin Binance coin vs Dogecoin Solana vs Dogecoin Cardano vs Dogecoin Polkadot vs Dogecoin
Bitcoin vs SHIBA INU Ethereum vs SHIBA INU Ripple vs SHIBA INU Binance coin vs SHIBA INU Solana vs SHIBA INU Cardano vs SHIBA INU Polkadot vs SHIBA INU Dogecoin vs SHIBA INU
Bitcoin vs Matic Network Ethereum vs Matic Network Ripple vs Matic Network Binance coin vs Matic Network Solana vs Matic Network