Talks
Speakers
Events
Topics
Search
Sign in
Search
Home
Talks
Speakers
Events
Topics
Leaderboard
Use
Analytics
Sign in
search talks for
⏎
Suggest modification to this talk
Title
Description
Neural networks (NNs) not only sound really cool, but they can also solve some pretty interesting problems ranging from driving cars to spam detection to facial recognition. Solving problems with NNs is challenging, because actually implementing a NN from scratch is difficult, and knowing how to apply it is more difficult. Fortunately, libraries, such as RubyFANN, exist to handle the first problem. Solving the second problem comes from experience. This talk will show a few different approaches to applying NNs to such problems as spam detection and games, as well as discussing other areas where NNs might be a useful solution. Help us caption & translate this video! http://amara.org/v/FG8g/
Date
Summarized using AI?
If this talk's summary was generated by AI, please check this box. A "Summarized using AI" badge will be displayed in the summary tab to indicate that the summary was generated using AI.
Show "Summarized using AI" badge on summary page
Summary
Markdown supported
In this talk at the LoneStarRuby Conf 2013, Ethan Garofolo introduces the concept of neural networks (NNs) and discusses their applications using the RubyFANN library. Key points covered in the presentation include: - **Understanding Neural Networks**: The speaker explains that neural networks are function approximators that use input data to find underlying patterns or rules, emphasizing the experimental nature of fine-tuning these models. - **Components of Neural Networks**: Each NN consists of input nodes, hidden neurons, and output nodes, with a focus on how these nodes interact during the training phase. - **Training Process**: Neural networks learn by adjusting the weights of connections based on errors encountered during processing, using algorithms like backpropagation to refine performance over time. - **Applicability of Neural Networks**: It is crucial to assess whether a neural network is suitable for a specific problem, with a warning against mistakes that can arise from improperly interpreting categorical data. - **Practical Applications**: Garofolo discusses spam detection as a case study, sharing his experience with data input variables and outputs when using RubyFANN to construct a spam detector. He notes that despite some success, neural networks may not be the optimal solution compared to other methods. - **Tic-Tac-Toe Example**: He also presents an example of analyzing game strategies using neural networks, contrasting different modeling approaches and highlighting the importance of data quality and learning contexts. - **Further Applications**: The speaker concludes by mentioning potential uses of NNs in facial recognition, handwriting recognition, and gesture detection, encouraging experimentation with various parameters and activation functions. Ultimately, Garofolo advises that while neural networks can be powerful, they require careful design and experimentation to effectively address real-world problems and underscores that every technical solution is fundamentally a human problem.
Suggest modifications
Cancel
