Log 6

 

For log 6, I want to talk about the practical source for machine learning. This is the first time I have actually discovered something useful and practical in this genre of books, which is why I am quite amazed and impressed. Most books only provide theories, information, datas and suggestions, as a result, we often stop discovering about that specific topic once we finish the book. However, this book not only gives out fundamental theory, but an actually existing and useful website for us to dive into the topic of our interest deeper and clearer with more practice and understanding. The website the book provides, Kaggle, is a site especially made for data science and machine learning. Kaggle hosts competitions in which data scientists and global companies compete to find out who can calssify or predict most efficiently, and the competition topics vary and change all the time. It's quite intersting because the winner's algorithms are featured on the site, meaning we can take a look at it, do a little research on it, and understand why the winner could do it the best way, thus learning from the masters of this field and makes Kaggle a really good resource for researchers interested in machine learning and data science. Moreover, it actually provides you with the data you need! If there is a topic in which you are interested, and the data is difficult to obtain, you can count on it. The data uploaded to Kaggle can be pre-tested to guide the overall direction and determine the progress of your research. The information it provides is quite abundant. For each competition it holds, Kaggle uploads a competition data set and a variety of data including text, images, and sound for researchers to test and analyze. It surprises me with the various types of information it offers, ranging from visual to aural data, the idea of analyzing an aural data makes me feel fascinated because we are so used to analyzing and sorting visual texts, data, etc. But what about sounds? I think we rarely hear about that, however, I find the idea of decoding the potential patterns in animals' sounds and trying to understand what ideas they are actually communicating engages my interest enormously, and I would love to explore this subtopic of machine learning more in the future.

I searched Kaggle on the internet and after browsing it and looking into it for a while, I found that the website offers more than the book told me. It was a huge and significant breakthrough for me! I found that it offers more than just simply data sets. It even offers machine learning models that are pre-trained and ready-to-deploy, they have models like Gemini by Google and Llama by Meta, and there are so many models that can do different works like audio command detection, speech-to-text, audio synthesis, and even text segmentation and image text recognition. I clicked into the Llama model by Meta and was astonished by how easy it  was to access the information and learn it. They would put the different versions of their model's code on the website and you can easily copy it and try it out yourself, with footnotes explaining what that specific line of code is for. I consider it quite easy for me to learn because I can understand each function and process of the model while it's analyzing data with the information provided on the website.

Besides, they provide free online courses! They have a comprehensive and gradually increasing in depth and complexity course design.  From introduction to programming with different subtopics and stages like from variables to functions, to data types and lists. Then after finishing this big topic, they allow you to go deeper to python and to introduction to Machine Learning. I especially adore the progressive learning journey, not to mention it divides them into smaller subtopics and sections so you don't feel too overwhelmed while trying to absorb the knowledge and feel the beauty of data science and machine learning. Last but not least, worrying about whether your computer can handle it? Absolutely no worries because the website got us all covered! Kaggle provides computing power like GPUs and TPUs that costs zero dollars and allows you to compoute and work in a powerful environment they call as "Kaggle Notebooks", and you can use it to write and run your code with an accelerator(GPU).

To conclude, Kaggle is really a powerful and useful platform for people who are interested in machine learning, data science and it greatly engages my interest. I will definitely look into it and try to learn something from the online courses and try out the cool feature of taking advantage of someone else's computing power instead of mine.