{
  "id": 522285,
  "title": "Bronze solution but an interesting insight [Genetic Programming]",
  "url": "/competitions/uspto-explainable-ai/discussion/522285",
  "author_name": "Viktoria",
  "post_date": "2024-07-25T11:48:45.014000",
  "votes": 8,
  "comment_count": 0,
  "views": 0,
  "content": "<p>The first place solution shows that classical computer science methods can outperform deep learning approaches.</p>\n<p>I was inspired by this <a href=\"https://www.kaggle.com/code/tubotubo/uspto-simulated-annealing-baseline\" target=\"_blank\">approach</a> and decided to stop using deep learning methods, but I didn't want to follow the most popular public solution, so I chose genetic programming. The tree-based approach was the most promising.<br>\nGP gave good results but required more resources than annealing, so I cooled down to this competition, but the second place shows that it was worth trying to rewrite the code in C++ :)</p>\n<p><a href=\"https://www.researchgate.net/figure/Illustration-of-the-genetic-programming-mutation-and-crossover-The-upper-left-expression_fig1_301846559\" target=\"_blank\">Example </a> of GP Tree approach for more clarity.</p>",
  "messages": [
    {
      "id": 2935546,
      "postDate": "2024-07-25T11:48:45.013Z",
      "content": "<p>The first place solution shows that classical computer science methods can outperform deep learning approaches.</p>\n<p>I was inspired by this <a href=\"https://www.kaggle.com/code/tubotubo/uspto-simulated-annealing-baseline\" target=\"_blank\">approach</a> and decided to stop using deep learning methods, but I didn't want to follow the most popular public solution, so I chose genetic programming. The tree-based approach was the most promising.<br>\nGP gave good results but required more resources than annealing, so I cooled down to this competition, but the second place shows that it was worth trying to rewrite the code in C++ :)</p>\n<p><a href=\"https://www.researchgate.net/figure/Illustration-of-the-genetic-programming-mutation-and-crossover-The-upper-left-expression_fig1_301846559\" target=\"_blank\">Example </a> of GP Tree approach for more clarity.</p>",
      "rawMarkdown": "The first place solution shows that classical computer science methods can outperform deep learning approaches.\n\nI was inspired by this [approach](https://www.kaggle.com/code/tubotubo/uspto-simulated-annealing-baseline) and decided to stop using deep learning methods, but I didn't want to follow the most popular public solution, so I chose genetic programming. The tree-based approach was the most promising.\nGP gave good results but required more resources than annealing, so I cooled down to this competition, but the second place shows that it was worth trying to rewrite the code in C++ :)\n\n[Example ](https://www.researchgate.net/figure/Illustration-of-the-genetic-programming-mutation-and-crossover-The-upper-left-expression_fig1_301846559) of GP Tree approach for more clarity.",
      "votes": 8
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2935546": "The first place solution shows that classical computer science methods can outperform deep learning approaches.\n\nI was inspired by this [approach](https://www.kaggle.com/code/tubotubo/uspto-simulated-annealing-baseline) and decided to stop using deep learning methods, but I didn't want to follow the most popular public solution, so I chose genetic programming. The tree-based approach was the most promising.\nGP gave good results but required more resources than annealing, so I cooled down to this competition, but the second place shows that it was worth trying to rewrite the code in C++ :)\n\n[Example ](https://www.researchgate.net/figure/Illustration-of-the-genetic-programming-mutation-and-crossover-The-upper-left-expression_fig1_301846559) of GP Tree approach for more clarity."
  }
}