Towards Data Science

It's a surprisingly difficult task with quite a bit of research history. We looked at different solutions extensively (https://medium.com/towards-data-science/editing-text-in-images-with-ai… and ended up building a tool to solve the problem: https://www.producthunt.com/posts/textify-2 Eventually models will get to the point where they can do this well natively but for now the best we can do is a post-processing step.

Good work! I did this project as well about a week ago. I will shamelessly also share my writings on this project: https://medium.com/towards-data-science/teaching-cars-to-see-vehicle-d… . The HOG + SVM method is quite slow and not as accurate as a deep learning approach. Before jumping onto semantic segmentation, I recommend re-implementing this project or more generally solve this problem using a Regional Convolution Neural Network architecture (R-CNN) like Faster R-CNN[1] or YOLO[2] for instance. [1]: https://arxiv.org/abs/1506.01497 [2]: https://arxiv.org/abs/1506.02640

Very interesting. I've implemented something similar[1] using a pure Collaborative Filtering approach[2][3], that I think works better for me, but it's unable to recommend unpopular repositories. The New York Times recommender system uses a hybrid approach (Content Based + Collaborative Filtering) called Collaborative Topic Modeling on top of LDA[4]. It would be interesting to try that. [1]: https://github-recs.appspot.com/ [2]: https://medium.com/towards-data-science/recommending-github-repositori… [3]: https://github.com/jbochi/facts [4]: https://open.blogs.nytimes.com/2015/08/11/building-the-next-new-york-t…

Great explanation! I also worked on a similar project (https://medium.com/towards-data-science/personality-for-your-chatbot-w… Also feel free to check my Conversation Analyzer (https://github.com/5agado/conversation-analyzer), which includes scraper and parser for Facebook Messanger conversations.

- Identify your goals. Are you blogging for extra income? Do you want to connect with a specific kind of reader, and grow a like-minded community? Do you want to be the best resource on the internet for a subject you love? Are you trying to get a new job? Will you write a book about this subject someday? Your goals inform how you'll promote your content. - Write THE authoritative post on a subject you are deeply passionate about. For me, this is "tiny machine learning," especially computer vision for single-board computers. A few of my posts are keyword-authoritative for "Raspberry Pi TensorFlow." patio11's salary negotiation post is a great example of an authoritative post. I've been linking my team-mates to this blog post 2x times a year since it was published, during raise/perf seasons. https://www.kalzumeus.com/2012/01/23/salary-negotiation/ - Publish on a consistent schedule. I struggle with this one, but every successful blog I know posts at least weekly. You'll have more hits the more times you step up to the plate. - Publish via multiple channels. I run a self-hosted Ghost blog & newsletter, and use Zapier to cross-post to medium.com. Make sure you provide a canonical URL, which tells search engines that your blog is the original source. - Submit guest posts to a popular publication (again, use a canonical link back to your primary domain) Example: https://medium.com/towards-data-science/3-ways-to-install-tensorflow-2… If you…

The link to the “new version” 404s, but this is probably a copy of it: https://medium.com/towards-data-science/the-inspection-paradox-is-ever… Also, this relates to renewal theory: https://en.wikipedia.org/wiki/Renewal_theory#Inspection_paradox

I once wrote this. https://medium.com/towards-data-science/embeddings-with-word2vec-in-no…

> Sounds a lot like BM25 weighted word embeddings (e.g. fastText). If you're interested in this topic, I wrote an article on this method back in 2020: https://medium.com/towards-data-science/building-a-sentence-embedding-…

This seems to explain the process: https://medium.com/towards-data-science/how-to-accelerate-and-compress…

This is an abstraction: https://medium.com/towards-data-science/hypertuning-for-tensorflow-pyt…

Not really. This is people looking at Wikipedia. Guess: Who's the Alan with the most pageviews in Wikipedia? Who's the Steve? I left the answers here: - https://medium.com/towards-data-science/bigquery-without-a-credit-card… Basically a show-off for BigQuery, but to answer this specific question: The most viewed Alan and Steve in Wikipedia are the ones closer to HN. SELECT title, SUM(views) views FROM `fh-bigquery.wikipedia_v3.pageviews_2019` WHERE DATE(datehour) BETWEEN '2019-01-01' AND '2019-01-10' AND wiki='en' AND title LIKE r'Alan\_%' GROUP BY title ORDER BY views DESC LIMIT 10

Could be better written but here are my thoughts on the same topic from earlier last year: https://medium.com/towards-data-science/hard-earned-advice-for-ai-prod…

"training with multiple altered versions of the image, e.g. various amounts of blur, noise and mean/median filters applied" "Are such techniques in use" Yes, it's called 'data augmentation': https://medium.com/towards-data-science/deep-learning-3-more-on-cnns-h…

You can read about it a bit here https://medium.com/towards-data-science/an-intuitive-guide-to-deep-net…

Nice. I wrote down my thoughts after taking his new DL course. Hope it helps you all :) https://medium.com/towards-data-science/thoughts-after-taking-the-deep…

Link is broken. Try this: https://medium.com/towards-data-science/symspell-vs-bk-tree-100x-faste…

Here is a good article: Hardware for Deep Learning https://medium.com/towards-data-science/hardware-for-deep-learning-8d9…

It's disappointing to see that the benchmark results are so opaque. I hope we see reproducible results soon, and hopefully from Mistral themselves. 1. We don't know what the evaluation setup is. It's very possible that the ranking would be different with a bit of prompt engineering. 2. We don't know how large each dataset is (or even how the metrics are calculated/aggregated). The metrics are all reported as XY.ZW%, but it's very possible that the .ZW% -- or even Y.ZW% -- is just noise.[1] 3. We don't know how the datasets were mined or filtered. Mistral could have (even accidentally!) filtered out particularly data points that their model struggled with. (E.g., imagine good-meaning engineer testing a document with Mistral OCR first, finding it doesn't work, and deducing that it's probably bad data and removing it.) [1] https://medium.com/towards-data-science/digit-significance-in-machine-…