week 21 Intro too AI

 Intro too AI


This week, we delved into the XOR challenge, a key puzzle in the field of artificial intelligence (AI) that initially stumped early AI models. The XOR (exclusive OR) logic gate is simple yet tricky: it outputs true only when the inputs differ. For a while, early neural networks couldn’t crack it due to its non-linear nature, highlighting a crucial limitation in AI's early days.

Addressing this puzzle required a rethink of neural network design. By integrating more layers and employing varied activation functions, these networks gained the ability to grasp the XOR’s complexities, marking a leap forward in AI's capability to solve non-linear problems.

Our practical exploration involved programming an Arduino Uno to tackle the XOR problem with a neural network. This setup included configuring the network with layers for input, hidden processing, and output, then training it with appropriate data. The training process is pivotal: through repeated cycles, the network adjusts its internal weights using a method called backpropagation, improving its prediction accuracy after each round.

Notably, the number of training cycles wasn't constant. Due to the random starting points set by initial weights and biases, each training session's duration to reach effective learning varied. This randomness is a natural part of neural network training, emphasizing the iterative process of learning and adapting.

Ultimately, the network reliably resolved the XOR challenge, showcasing AI’s prowess in navigating complex problems. This not only demonstrates AI's potential in addressing intricate issues but also the critical role of iterative refinement and the experimental spirit in advancing AI technology.

The journey through the XOR problem illuminates a pivotal moment in AI history, reflecting on how neural networks have evolved to overcome complex challenges through innovation and continuous learning. It's a testament to the importance of training cycles in the AI development process, underscoring the blend of theoretical knowledge and practical experimentation in advancing the field.





We're creating a simple neural network with three parts: an input section, a hidden section, and an output section. This setup is similar to how a digital display works, where giving it a number between 0 and 9 shows that exact number on the screen.






Comments

Post a Comment

Popular posts from this blog

Week 11-17 Group Project

Image
  In our group project, my friends Shakline, Nelushan, and I teamed up to create something exciting: a lie detector machine. Our aim was to use Arduino technology to figure out when someone might be lying. We used two fancy sensors: one to measure heart rate changes (ECG sensor) and another to track sweat levels on the skin (GSR sensor). These changes often happen when someone is feeling stressed, like when they're not telling the truth. To complete our lie detector setup, we also placed the skin sensor between two fingers. This sensor helps measure changes in skin conductivity, which can indicate stress levels. By positioning it this way, we could get accurate readings of how much a person might be sweating, adding another layer to our detection system. For our lie detector, we placed the pads of the ECG sensor on different pulse points. We used three colors of pads: green, yellow, and red. Each color was for a specific pulse point. We had a picture showing where to put the pads ...
Read more

week 18 Recommender systems

Image
  Recommender systems Exploring Recommender Systems: Shaping User Experiences This week's exploration led me into the fascinating world of recommender systems, which play a vital role in shaping the user experiences of modern online platforms like Netflix and Amazon. Throughout this journey, I delved into understanding how these systems personalize recommendations based on individual preferences derived from past interactions. Harnessing Data for Personalization Recommender systems harness various data inputs, including user behavior and item characteristics, to suggest products or content tailored to each user. Beyond enhancing user experience, these systems are essential for businesses to stay competitive by quickly adapting to consumer needs and staying ahead of technological advancements. The Role of Machine Learning in Recommender Systems A key aspect of our exploration was uncovering the role of machine learning in powering recommender systems. Machine learning algorithms an...
Read more