What does supervised learning mean?

Supervised learning is one of the three main categories in machine learning: supervised learning, unsupervised learning, and reinforcement learning.

In supervised learning, the algorithm works with labeled training data, which means that the input data (features) are paired with the corresponding output data (labels or target variables). The goal is to learn a function or rule that captures the relationship between the input data and the output data.

Your contact person.

Agnes-Tholen-Companymind-Small

Agnes Tholen

Director Sales & Marketing (ppa.)

What is Supervised Learning?

Let’s imagine we want to use a supervised learning model in the food industry to detect the ripeness of fruit. For this purpose, we collect a large amount of images of fruits in various degrees of ripeness. The images already have the appropriate labels indicating the degree of ripeness of the fruit, for example, “unripe”, “ripe” or “overripe”.

The training data is provided to the algorithm so that it can learn from the examples. During the training process, the algorithm analyzes the images and learns to recognize characteristic features and patterns that correlate with the different degrees of ripeness of the fruit.

For example, the model learns that unripe fruits have a green color and a firm texture, while ripe fruits have a bright color and a softer texture. Overripe fruit, on the other hand, might have brown spots or wrinkly skin.

After the model has been sufficiently trained, it can be applied to new images of fruit it has never seen before. The model can then predict the ripeness of the fruit based on the learned characteristics.

There are many other exciting applications in the world of supervised learning, such as recognizing handwritten text, classifying emails as spam or not spam, predicting sales figures, and much more. It is a powerful way to develop intelligent solutions:

Opengraph

Application areas for Supervised Learning are:

  • Image recognition: classification of images into different objects or categories.

  • Speech processing: classification of texts, speech translation or speech recognition.

  • Medical diagnosis: Prediction of diseases based on medical images or patient data.

  • Finance: Predicting financial markets or credit risk.

  • Autonomous vehicles: classification of traffic signs or obstacles.

Back to the overview

The main types of tasks in supervised learning

Classification

Here, the algorithm is trained to classify input data into various predefined classes or categories. One example is the classification of e-mails into “spam” or “non-spam”.

Regression

Regression is about predicting continuous output values. The algorithm learns to create a function that maps the input data to a continuous target variable. One example is predicting the selling price of a property based on its characteristics.

Application examples in the food industry

One application example of how supervised learning works in the food industry is the classification of foods based on their ingredients or properties.

Imagine a food manufacturer wants to develop an automated system that classifies food products into different categories based on their ingredients, such as “vegetarian,” “vegan,” “gluten-free,” or “lactose-free.” The goal is to give consumers quick and accurate information about products without having to manually read every label.

To create this system, the manufacturer collects an extensive database of food products that are already labeled accordingly. Each record contains the list of ingredients and the corresponding categories (e.g. vegetarian or vegan).

These data serve as training data for the supervised learning algorithm. The model analyzes the data and learns which combinations of ingredients typically lead to certain categories. It can recognize that products with meat as an ingredient are not vegetarian or vegan and that products with dairy ingredients are not lactose-free.

After training, the model can be used to automatically classify new food products. When a new product hits the market, the system can determine which category it belongs to based on its ingredients and display that information on the label or make it available online.

By using supervised learning, food manufacturers can quickly and reliably provide information about their products, helping consumers make informed decisions about their diets. It is an example of how machine learning can help improve transparency and consumer convenience in the food industry.

Predicting sales volumes for food products in supermarkets is also feasible thanks to supervised learning: suppose a food manufacturer or retailer wants to predict the sales volume for a particular product, such as breakfast cereal, in its stores. To do this, it collects historical sales data of this product over a certain period of time, including sales figures and the respective features such as promotions, special offers, days of the week, holidays, weather conditions, etc. This data serves as training data for the supervised learning algorithm.

The goal is to build a model that learns the relationship between input attributes (promotions, weather, etc.) and outputs (sales figures) to predict future sales figures.

During the training process , the algorithm analyzes the data and looks for patterns and correlations between the input attributes and the sales figures. For example, it may find that sales are higher on sunny days, when promotions are running, or when the product is offered at a special price.

After the model is well trained, it can be used to make sales predictions for new input attributes. Suppose the retailer is planning a promotion for the product and wants to know how sales are likely to change. The model can provide an estimate of the expected sales volume based on the input attributes.

By using supervised learning, companies in the food industry can better inform their business decisions and make them more efficient. In this case, sales volume forecasting enables better planning of inventory levels and production quantities to avoid bottlenecks or overstocks.

Back to the overview

What is Unsupervised Learning?

Unsupervised learning is another important category in machine learning. Unlike Supervised Learning, where the algorithm uses labeled training data and the correct answers are predetermined, Unsupervised Learning uses unlabeled data where the categories or classes are not known. The main goal of Unsupervised Learning is to identify patterns and structures in the data in order to group, segment or reduce them. So it’s about extracting hidden or intrinsic information from the data.

The two main types of tasks in Unsupervised Learning

1 – Clustering

Clustering algorithms group similar data points into clusters, with members of one cluster having more similar characteristics than those of other clusters. These algorithms make it possible to detect natural groupings in the data. An example is the k-means algorithm, which divides data into k clusters, where k is specified beforehand.

2 – Dimensionality Reduction (Dimensionality Reduction)

This uses techniques to reduce the number of features or dimensions in the data while retaining the relevant information. This can be helpful to improve data visualization or speed up the training process. One example is principal component analysis (PCA).

Your contact person.

Agnes-Tholen-Companymind-Small

Agnes Tholen

Director Sales & Marketing (ppa.)

Opengraph

Application Areas for Unsupervised Learning:

  • Image recognition: classification of images into different objects or categories.

  • Speech processing: classification of texts, speech translation or speech recognition.

  • Medical diagnosis: Prediction of diseases based on medical images or patient data.

  • Finance: Predicting financial markets or credit risk.

  • Autonomous vehicles: classification of traffic signs or obstacles.

Back to the overview

What is reinforcement learning?

For a successful delimitation of the three main categories, reinforcement learning is still missing:

Reinforcement learning is a learning approach in which an agent learns how to perform actions to achieve a specific goal through interaction with an environment. The agent is confronted with rewards or punishments for his actions, which motivates him to develop optimal strategies to achieve long-term goals.

Optimizing food production with reinforcement learning

Imagine you have a factory that produces snack chips. In this factory, various steps are performed, such as cutting, frying, seasoning and packaging the chips. The goal is to maximize production output and ensure product quality while minimizing energy consumption. This is where reinforcement learning comes in:

Agent

The “agent” in this case would be a software program or AI that has control over the production processes and can make decisions to achieve the desired goals.

Actions

The agent can perform various actions, such as adjusting the frying temperature, changing the cutting speed or adjusting the seasoning amount.

Environment

The environment is the food factory itself, where the production takes place. It provides the agent with information about the current state of the processes, production performance, chip quality and energy consumption.

Reward

The agent receives a reward after each action based on the factory’s performance. Positive rewards can be given for higher production output and quality, while negative rewards are given for energy waste or inferior products.

In food production and distribution, efficient warehousing and logistics are critical to ensure that fresh and perishable foods reach their destinations on time.

Supply Chain Optimization with reinforcement learning

Reiforcement Learning enables automated and intelligent supply chain optimization to improve efficiency, resource utilization, and delivery accuracy.

Agent

In this case, the agent can be a system that makes inventory and transportation decisions.

Actions

The agent can perform actions such as ordering food, assigning storage locations, planning transportation routes, and controlling delivery vehicles.

Environment

The environment includes the entire supply chain, from food production to warehouses to retail outlets. Information such as inventory levels, demand forecasts, delivery schedules, traffic conditions, and weather conditions are all part of the environment.

Reward

Rewards can be awarded based on factors such as on-time delivery, minimizing spoilage, optimizing transportation costs, and avoiding overstocks or shortages.

The reinforcing learning algorithm analyzes the environment and makes decisions to optimize inventory and transportation to achieve desired goals. The system learns over time which decisions lead to the best results and adjusts its strategies accordingly. This can help reduce food waste, shorten delivery time, lower storage costs, and increase the overall efficiency of the food supply chain.