Decoding ML - From Basic Concepts to Complex Challenges

Decoding ML - From Basic Concepts to Complex Challenges

Welcome back to the second article in this engaging series on AI, Machine Learning, and Generative AI. After providing a broad overview in our initial piece, “A Br(AI)ve New World”, we now delve deeper into the world of Machine Learning, distinguishing it from traditional programming, and exploring its mechanisms and diverse applications. This journey will take us beyond the basics and into the application of these concepts in a meaningful, impactful way. Let’s continue our exploration of these transformative technologies together. Exciting times ahead, so stay tuned! 🚀🧠💡

What is the difference between machine learning and classical programming?

What’s the difference between machine learning and traditional programming, you ask? Well, with machine learning, it’s a bit like teaching a computer to find its own way round a maze, using examples as breadcrumbs. It doesn’t need a set map, it just learns to spot patterns in the maze (or data). In other words, machine learning is about crafting algorithms that pick up on these patterns and make educated guesses based on them. It’s a bit like a detective getting sharper with every case they solve. This learned wisdom can then be applied to new, unseen data, without the need for any new coding. In essence, with machine learning, you start with a mystery, find clues (data), use your ‘algorithm detective’ to make sense of it all, and then crack the case (produce an output). Traditional programming, on the other hand, requires a much more by-the-book approach, laying out specific rules for every scenario.

How does machine learning work?

How does machine learning go about its business, then? Well, there are three main ways it gets the job done, depending on the task at hand.

Supervised learning is a bit like a student learning with a textbook. Here, a model gets taught using a data set which has the correct answers already paired with the questions. It’s about learning the right answers for the test.

Unsupervised learning is more like learning to ride a bike without stabilisers. Here, the training model gets stuck in with the data without any hand-holding. The aim of the game here is to spot patterns and structures all on its own.

Lastly, there’s reinforcement learning. This method is a bit like training a dog – the training model learns from its environment by getting a pat on the back for good behaviour and a telling off for the wrong ones.

What kind of solutions can ML provide?

Machine Learning is a powerful tool capable of delivering a diverse array of solutions. These include forecasting outcomes (such as regression and classification), sorting tasks (like rankings and scores), and discerning tendencies (through recommendations and clustering).

• Regression: This involves determining a numeric value. An application of this might be estimating competitive prices for commodities.
• Classification: This task involves assigning a category or label. An example could be marking fake products to protect consumers and businesses.
• Ranking: Here, Machine Learning assists in arranging entities to spotlight the most pertinent ones.
• Recommendation: This involves pinpointing items of interest based on previous interactions or behavior, helping tailor experiences and offerings.
• Clustering: This allows grouping of entities, such as customers, to discern different market segments. This approach aids in identifying patterns and trends in various examples.
• Anomaly Detection: This method allows for the identification of outliers or anomalies within given examples, such as pinpointing products that have been miscategorized.

What are some potential problems with machine learning models?

Machine learning modelling can be problematic for learning algorithms due to the ingestion of poor-quality data. For example, the data may not include enough samples to represent a sufficiently broad scope of relevant variables.

What are simple and complex models?

Simple and complex ML models differ when balancing a model’s accuracy (number of correctly predicted data points) and a model’s explainability (how much of the ML system can be explained in “human terms”). The output of a simple ML model may be explainable and produce faster results, but the results may be inaccurate. The output of a complex ML model may be accurate, but the results may be difficult to communicate.

What is unexplainabilty?

Unexplainability refers to the extent to which we cannot readily articulate or understand the logic behind a Machine Learning model’s decisions. The acceptability of unexplainability can change depending on various circumstances and could be influenced by legal, ethical, professional or regulatory considerations.

When might it be a problem?

• If you have to explain to a customer why their loan application was turned down.
• If you need to justify why a particular transaction was flagged as suspicious or fraudulent.

When might it be acceptable?

• If the consequences of a misjudgement are minimal, such as when identifying objects for a catalogue search, or predicting the likelihood of an American football play being completed successfully.
• If it’s ultimately people, rather than algorithms, who have the final say on a decision.

What is uncertainty?

In plain terms, uncertainty refers to an outcome that’s not quite spot on. When we’re chatting about machine learning, uncertainty comes into play when we’re dealing with these clever models of ours.

You see, these models are a bit like detectives. They try to fit together all the pieces of the puzzle using training data, which can be a bit like trying to complete a jigsaw with a few missing pieces. This data they’re using might not be perfect – it could be a bit skew-whiff, a bit off, or even downright dodgy.

What’s more, the elusive ‘best’ data – the gold standard, the bee’s knees of all data – might be as hard to pin down as a greased pig at a country fair. So, all these elements can lead to a bit of uncertainty in the results, as your machine learning model does its best to make sense of what it’s got.

As I wrap up this second deep-dive into the world of AI, I hope you’re beginning to get your head around the complex and exciting field of Machine Learning. We’ve journeyed together from the very basics to the meatier aspects, explaining how it ticks and what it might mean for us in the day-to-day. And trust me, we’ve only just scratched the surface! Our goal at Parsectix is to guide you through this thrilling journey, deciphering all the AI and ML terminology as we go. I believe that knowledge is power and in the world of AI, it’s the power to drive change and spark innovation.

Keep a lookout for the next instalment in this series as we continue to explore the intriguing world of AI together. Until then, let’s keep riding this 🌊 of technological transformation.

If you happened to miss the first article, do pop back and have a read here: “A Br(AI)ve New World”

‍