Artificial Intelligence Lecture Notes PDF

# Artificial Intelligence ## Model of an artificial neuron The image shows a diagram illustrating the model of a basic artificial neuron. - The neuron receives inputs **x₁**, **x₂**, ..., **x_n**. - These inputs are weighted by **w₁**, **w₂**...

# Artificial Intelligence ## Model of an artificial neuron The image shows a diagram illustrating the model of a basic artificial neuron. - The neuron receives inputs **x₁**, **x₂**, ..., **x_n**. - These inputs are weighted by **w₁**, **w₂**, ..., **w_n**. - The weighted inputs are summed up to calculate **net**. - **net** is then passed through an activation function **f** to produce the output **y**. **Terminology:** 1. **Inputs:** **x₁**, **x₂**, ..., **x_n** are the inputs to the neuron. 2. **Weights:** **w₁**, **w₂**, ..., **w_n** are real-valued parameters called weights. 3. **Weighted sum:** **net = w₁ x₁ + w₂ x₂ +...+ w_n x_n** is the weighted sum of the inputs. 4. **Activation function:** **f** is called the activation function. It determines the output of the neuron based on the weighted sum. 5. **Output:** **y = f(net)** is the output of the neuron. ## Network Architecture ### Single layer net The image shows a simple neural network with a single layer. It consists of an input layer, an output layer, and a set of connections between them. ### Multi-layer net This diagram depicts a multi-layer neural network. It has an input layer, an output layer, and multiple hidden layers between them. - The input layer receives data from the outside world. - The hidden layers perform complex calculations on the input data. - Finally, the output layer produces the final result. ## Offline Learning This image shows a schematic representation of offline learning in artificial intelligence. - The process begins with training a brain network using data from healthy subjects. - This initial model is then used as a prototype for **SSAE (Sparse Stacked Autoencoder)**, which is a type of neural network used for feature learning. - Subsequently, the model is retrained using data from targeted subjects, incorporating prior knowledge to improve its performance. - This leads to the development of a specialized **SSAE** model. - Finally, a **DTL-NN (Deep Transfer Learning Neural Network)** is trained using the specialized **SSAE** and performs classification on a new subject. ## Supervised Learning The image illustrates the concept of supervised learning in machine learning: - **Supervised Learning:** Training data is labeled, providing the algorithm with information about the desired output for each input. This allows the algorithm to learn a relationship between the inputs and outputs. - **Examples:** The example shows images of cats and dogs with labels assigned to them. The AI algorithm would be trained using this data to learn to correctly identify cats and dogs in new images. ## Unsupervised Learning The image explains unsupervised learning in machine learning: - **Unsupervised Learning:** In unsupervised learning, the data provided to the algorithm is not labeled. The algorithm must discover patterns and relationships within the data without any guidance on the expected output. - **Examples:** The image shows various unlabeled items. The task for the algorithm is to learn about them and categorize them based on their characteristics. ## Machine Learning Machine learning is a field of computer science that focuses on developing algorithms that can learn from data and make predictions. These algorithms learn by building models from example inputs, enabling them to make data-driven decisions rather than following static instructions. ## Types of Machine Learning Two main types of machine learning are discussed: - **Supervised Learning:** The algorithm is trained on labeled data with known outputs. - **Unsupervised Learning:** The algorithm explores unlabeled data to discover patterns and structures. ## What is Supervised Learning? The image depicts an analogy to explain supervised learning: - Imagine having separate baskets for various fruits (like yellow bananas, golden pineapples, and black grapes). - When you are presented with a golden pineapple, you know exactly where to put it because you have learned the association between the fruit and its corresponding basket. - Supervised learning is similar. The algorithm is trained on labeled data, establishing a relationship between inputs and outputs. ## The Supervised Learning Process The image outlines the supervised learning process: - **Training Data:** The algorithm is trained using labeled data. - **Machine Learning Algorithm:** The algorithm learns a relationship between the inputs and outputs based on the training data. - **Predictive Model:** The trained algorithm creates a model that can predict outputs for new inputs. - **Model Evaluation:** This step assesses the accuracy of the predictive model. - **Feedback Loop:** The model can be further refined and improved through feedback, utilizing new data and adjusting the algorithm. ## What is Unsupervised Learning? The image provides an analogy to illustrate unsupervised learning: - Imagine being given 100 unlabeled items. After becoming familiar with these items, you are then given a new item. - You might be able to confidently determine what the new item is solely based on your understanding of similar unlabeled items you have encountered before. - This is the core concept of unsupervised learning. The algorithm learns patterns and structures in the data without any prior labeling. ## The Unsupervised Learning Process The image depicts the unsupervised learning process: - **Input Data:** The algorithm receives unlabeled data as input. - **Machine Learning Algorithm:** The algorithm analyzes the data to identify hidden patterns and structures. - **Output:** The algorithm produces an output based on the discovered patterns and relationships. Unsupervised learning is computationally complex and often less accurate than supervised learning but is useful for tasks such as clustering and dimensionality reduction.