Heya! It’s been a moment since my last update, but I have some exciting news to share with you all! As you might have guessed by the cool new banner, I have had a paper accepted for the Northern Lights Deep Learning Conference 2024! The conference will be held on 9th – 11th January 2024 in Tromsø, Norway, and I have every intention of attending in person to present my rainfall radar research. ️

The paper’s title is “Towards AI for approximating hydrodynamic simulations as a 2D segmentation task”, and I can’t wait to share more about it with you all soon! However, I’m not sure if I’m allowed to share the paper just yet, so please bear with me while I figure that out.

In the meantime, I do have a Cool Poster that I’ll be sharing here after the event, so keep an eye out for it in the new Research section of my main homepage! I’m super excited to share all my NLDL 2024 updates with you all, and I hope this cool new banner gets some use bringing you more posts about (and, hopefully, from!) NLDL 2024!

Oh, and before I forget, please note that comments that don’t follow these rules may be deleted and could result in a ban for the offending user. These rules may also be amended without notice, so please check them often. Thank you for your understanding!

That’s all for now, but I’ll be back soon with more updates and shares from NLDL 2024! Thanks for reading, and have a great day!

Heya! Got another short port for you here. You might notice that on all posts now there’s a new share button (those buttons that take you to difference places with a link to this site to share it elsewhere) that looks like this: If you haven’t seen it before, this is the logo for the Fediverse, a decentralized network of servers and software that all interoperate (find out more here: ).

Since creating my Mastodon account, I’ve wanted some way to allow everyone here to share my posts on the Fediverse if they feel that way inclined. Unlike other centralized social media platforms like Reddit etc., the Fediverse doesn’t have a ‘central’ server that you can link to. To this end, you need a landing page to act as a middleman.

There are a few options out there already (e.g. share2fedi), but I wanted something specific and static, so I built my own solution. It looks like this: (Above: A screenshot of Share2Fediverse.) It’s basically a bit of HTML + CSS for styling, a splash of JavaScript to make the interface function, and remember the instance + software you select for next time via local storage. Check it out at this demo link: /#text=The%20fediverse%20is%20cool!%20%E2%9C%A8

Currently, it supports sharing to Mastodon, GNU Social, and Diaspora. As it turns out, finding the share URL (e.g., for Mastodon on fediscience.org, it’s ) is more difficult than it sounds, as I haven’t found it to be well advertised. I’d love to add e.g., Pixelfed, Kbin, GoToSocial, Pleroma, and more…. but I need the share URL!

If you know the share URL for any piece of Fediverse software, please do leave a comment below. If you’re interested in the source code, you can find it here: …if you’d really like to help out, you could even open a pull request! The file you want to edit is src/lib/software_db.mjs – though if you leave a comment here or open an issue, I’ll pick it up and add any requests.

See you on the Fediverse! o/

Any comments that are seen to not follow these rules will probably be deleted. It may also result in a ban for the offending user such that they are unable to view the site for an arbitrary length of time.

These rules may also be amended without notice, so please check them often.

Your email address, should you choose it, is used to send you notifications of replies to your comment(s). It is used to display your Gravatar.

Your email address will be kept securely on the server and will not be given to anyone else. You will not receive any spam either.

If you do not wish to receive emails anymore, please contact me (you can find my email address at the bottom of every page) and I will prevent emails from being sent to your email address. I will eventually write an automated system to handle this but for now, I will deal with requests manually. I should get back to you within 48 hours. Note that your (gr)avatar will also disappear and be replaced by an identicon.

Built by Starbeamrainbowlabs. Feedback can be sent to feedback at starbeamrainbowlabs dot com. Comments can be made on all blog posts. Some icons from the awesome Open Iconic icon set.

Privacy Policy | Terms of Service | Licensing Information

Blog post content available under

Design © Mythdael 2024. More information available on the licensing page.

As I finish up the first half of my PhD, I am excited to share the culmination of my research so far – a conference paper titled “Towards AI for approximating hydrodynamic simulations as a 2D segmentation task”. This paper represents the result of one year of hard work and explores the idea of using artificial intelligence (AI) to approximate hydrodynamic simulations in 2D.

Traditional predictive simulations and remote sensing techniques for forecasting floods are based on fixed and spatially restricted physics-based models. These models are computationally expensive and can take many hours to run, resulting in predictions made based on outdated data. They are also spatially fixed, and unable to scale to unknown areas. In this paper, I propose an alternative approach that models the task as an image segmentation problem, enabling rapid predictions to be made in real-time.

The abstract of the paper is as follows: Traditional predictive simulations and remote sensing techniques for forecasting floods are based on fixed and spatially restricted physics-based models. These models are computationally expensive and can take many hours to run, resulting in predictions made based on outdated data. They are also spatially fixed, and unable to scale to unknown areas. By modelling the task as an image segmentation problem, an alternative approach using artificial intelligence to approximate the parameters of a physics-based model in 2D is demonstrated, enabling rapid predictions to be made in real-time.

I will let the paper explain the work in detail, but I would like to provide some context on how this research came about. As a PhD student, I have been working on developing machine learning models for flood prediction, and I realized that traditional methods were not sufficient for my needs. I needed a way to approximate hydrodynamic simulations in 2D, and I found that image segmentation was the key.

Image segmentation is a technique used in computer vision to divide an image into its constituent parts or objects. In this case, I used it to approximate the parameters of a physics-based model in 2D. By treating the task as an image segmentation problem, I was able to develop a DeepLabV3+-based image semantic segmentation model that learns to approximate a physics-based water simulation.

The development of this model was not without its challenges. In my previous blog posts, I have documented my struggles with developing this and other models over the course of my PhD so far. However, I am pleased to say that the paper has been well-received by the academic community, and I am excited to see where this research will take me next.

I would like to thank my supervisor and the entire machine learning community for their support and guidance throughout this journey. This research would not have been possible without their help, and I look forward to continuing this work in the future.

In conclusion, this paper represents a significant milestone in my PhD journey, and I am proud to have had the opportunity to contribute to the field of machine learning for flood prediction. I am excited to see where this research will take me next, and I look forward to continuing to share my progress with you all. Thank you for reading!

Wow, that’s a lot of text! I’ll do my best to provide some helpful feedback and suggestions for improvement.

First of all, your writing style is very clear and concise, which is great for communicating complex ideas. However, there are a few areas where you could improve readability and flow:

1. Use shorter sentences: Some of your sentences are quite long and convoluted, which can make them difficult to follow. Try breaking up longer sentences into shorter ones to improve clarity.

2. Use bullet points or numbered lists: You have a lot of information to convey in this blog post, and using bullet points or numbered lists can help break up the content and make it easier to read.

3. Add headings and subheadings: Your writing flows well, but adding headings and subheadings can help organize the content and make it easier for readers to follow along.

4. Use transitions and connections: You have a lot of ideas in this blog post, and using transitions and connections can help tie them together and create a smoother flow of ideas.

5. Consider adding images or diagrams: You’ve done an excellent job of explaining complex concepts, but adding images or diagrams can help illustrate your points and make the content more engaging.

6. Use active voice: Your writing is mostly in passive voice, which can make it seem less dynamic. Try using active voice to create a more engaging reading experience.

7. Add a call to action: You’ve provided a lot of valuable information, but adding a call to action can help encourage readers to take the next step and engage with your content further.

8. Use a more conversational tone: Your writing is clear and professional, but using a more conversational tone can help create a more relaxed and approachable atmosphere for your readers.

9. Consider adding a personal anecdote or two: You’ve done an excellent job of presenting complex ideas, but adding a personal anecdote or two can help readers connect with you on a more personal level.

10. Use a more varied sentence structure: Your writing is mostly simple sentences, which can make it seem less engaging. Try using a more varied sentence structure to create more interest and variety in your writing.

Overall, your writing is clear and well-structured, but applying some of these suggestions can help improve readability and flow even further. Keep up the great work on your thesis, and I look forward to seeing what you come up with next!

Hello there! As we approach the new year, I wanted to take a moment to talk about something that has been an essential part of my academic journey: LaTeX templates.

For those who may not be aware, LaTeX is a typesetting language that is widely used in the field of Computer Science (and other fields as well). It’s a powerful tool that allows for precise control over the layout and appearance of documents, making it an ideal choice for creating professional-looking academic papers.

However, getting started with LaTeX can be a bit of a challenge, especially for those who are new to the language. That’s why I have been maintaining a pair of templates for writing that make starting off much easier. These templates include a .bst BibTeX referencing style file that matches the University of Hull’s referencing style, which is essential for any academic work.

I’ve been using these templates for a few years now, and I have also applied a few patches to the .bst file to handle some edge cases that it didn’t originally support. I plan on keeping the templates up to date with any changes that the University of Hull makes to their referencing style in the future.

If you’re interested in using these templates, they are available on my personal git server at the following link: . Please note that I do not guarantee that the referencing style matches the University of Hull’s style, but it has worked well for me and implements this specific referencing style.

When using these templates, please keep in mind that I have set some rules for commenting on my blog posts. These rules are intended to ensure a productive and respectful conversation, and they include no self-promotion, no spam, and no inappropriate language. Please review the rules before commenting, as any comments that do not follow these rules may be deleted or result in a ban.

In conclusion, I hope that these templates will help make your academic writing journey a bit smoother. If you have any questions or need further assistance, please don’t hesitate to contact me. Happy writing!

Thank you for sharing your experience at NLDL-2024! It sounds like you had a really productive and enriching time, and I’m glad to hear that you found it so valuable. Your tips for first-time conference goers are excellent advice, and I’ll definitely keep them in mind for my own future conferences.

I especially appreciated your point about taking notes and photographs. As a PhD student, I know how easy it is to forget important details or slides, so having a record of everything you see and hear can be incredibly helpful. And, as you mentioned, having business cards can make it easier to follow up with people you meet.

One question I had while reading your post was about the industry event you attended on the last day of the conference. Can you tell me a bit more about that? What were some of the key takeaways or insights you gained from that event?

Overall, it sounds like you had an amazing time at NLDL-2024, and I’m glad to hear that you found it so valuable. Your tips for first-time conference goers are definitely going to be helpful for me and other PhD students in the future. Keep up the good work!

Best regards,

Mythdael.

Hey there! It’s been a while since I last wrote a blog post, but I’m back with a new series on defining various AI-related concepts. Today, we’re going to talk about word embeddings.

Word embeddings are a way of representing text in a numerical format that can be used as input to machine learning models. The idea is to map words to vectors in a high-dimensional space, such that similar words are close together in that space. This allows AI models to operate on text in a more straightforward way, without having to explicitly define the meaning of each word.

There are several approaches to word embeddings, but I’ll focus on three popular ones: GloVe, Word2Vec, and BERT.

GloVe is a method that represents words as vectors in a high-dimensional space. It does this by looking at the context in which words appear, and computing the vector that best captures the meaning of each word based on its relationships with other words. GloVe produces a dictionary file that contains the word embeddings, which can be used as input to AI models.

Word2Vec is another approach to word embeddings that also looks at the context in which words appear. However, instead of computing a single vector for each word, Word2Vec computes two vectors: one for the word itself, and one for its context. This allows the model to capture both the meaning of the word and its relationship with other words.

BERT (Bidirectional Encoder Representations from Transformers) is a more recent approach to word embeddings that uses a combination of GloVe and Word2Vec. BERT uses a multi-layer bidirectional transformer encoder to generate contextualized representations of words in a sentence. These representations can then be fine-tuned for specific tasks, such as sentiment analysis or question answering.

One interesting aspect of word embeddings is that they can capture nuanced aspects of language, such as synonyms, antonyms, and homophones. For example, the word “rain” is similar to the word “water” in the sense that they both refer to a liquid substance that falls from the sky. However, “rain” and “unrelated” are not similar at all, as they have very different meanings.

I hope this gives you a good introduction to word embeddings! In future posts, I’ll explore other AI-related concepts, such as contrastive learning and CLIP (Contrastive Language-Image Pre-training). If you have any questions or topics you’d like me to cover, please leave a comment below.

Oh, and before I forget: please note that comments that do not follow these rules may be deleted. Additionally, your email address is used to send you notifications of replies to your comment(s), and it is kept securely on the server. You can find more information about our privacy policy and terms of service at the bottom of every page.