{
  "id": 174834,
  "title": "Image analysis with TSNE and KMeans",
  "url": "/competitions/digit-recognizer/discussion/174834",
  "author_name": "DeepUnderstanding",
  "post_date": "2020-08-15T17:05:06.271000",
  "votes": 0,
  "comment_count": 0,
  "views": null,
  "content": "<p>\"Let the neural network sleep\", recently I have come out with this statement in some article. The article said that the trend of just making an end to end neural network model without understanding the data is increasing creating many others and mysteries of the data unsolved.<br>\nSo simply put, here I have tried to analyse the Image data provided. I have used a famous Dimension reducing algorithm called TSNE to make each image fit to a cartesian coordinate system, that is each image is represented as a point having x and y coordinate. <br>\nAn image is a multidimensional data hence very difficult to understand and to use simple machine learning techniques, with TSNE we have reduced the dimension and can see how each image is related to one another.<br>\nIn the end, I have used a simple clustering technique called Kmeans to segregate the different data points and compare it with the original.<br>\nThe related notebook is <a href=\"https://www.kaggle.com/mrinath/image-analysis-with-tsne-and-kmeans?scriptVersionId=40804344\" target=\"_blank\">here</a></p>\n<p>To know more about TSNE you can watch <a href=\"url\" target=\"_blank\">https://www.youtube.com/watch?v=NEaUSP4YerM</a><br>\n and you can find more on Kmeans <a href=\"https://www.youtube.com/watch?v=4b5d3muPQmA\" target=\"_blank\">here.</a>Thank You!!</p>",
  "messages": [
    {
      "id": 971555,
      "postDate": "2020-08-15T17:05:06.273Z",
      "content": "<p>\"Let the neural network sleep\", recently I have come out with this statement in some article. The article said that the trend of just making an end to end neural network model without understanding the data is increasing creating many others and mysteries of the data unsolved.<br>\nSo simply put, here I have tried to analyse the Image data provided. I have used a famous Dimension reducing algorithm called TSNE to make each image fit to a cartesian coordinate system, that is each image is represented as a point having x and y coordinate. <br>\nAn image is a multidimensional data hence very difficult to understand and to use simple machine learning techniques, with TSNE we have reduced the dimension and can see how each image is related to one another.<br>\nIn the end, I have used a simple clustering technique called Kmeans to segregate the different data points and compare it with the original.<br>\nThe related notebook is <a href=\"https://www.kaggle.com/mrinath/image-analysis-with-tsne-and-kmeans?scriptVersionId=40804344\" target=\"_blank\">here</a></p>\n<p>To know more about TSNE you can watch <a href=\"url\" target=\"_blank\">https://www.youtube.com/watch?v=NEaUSP4YerM</a><br>\n and you can find more on Kmeans <a href=\"https://www.youtube.com/watch?v=4b5d3muPQmA\" target=\"_blank\">here.</a>Thank You!!</p>",
      "rawMarkdown": "\"Let the neural network sleep\", recently I have come out with this statement in some article. The article said that the trend of just making an end to end neural network model without understanding the data is increasing creating many others and mysteries of the data unsolved.\nSo simply put, here I have tried to analyse the Image data provided. I have used a famous Dimension reducing algorithm called TSNE to make each image fit to a cartesian coordinate system, that is each image is represented as a point having x and y coordinate. \nAn image is a multidimensional data hence very difficult to understand and to use simple machine learning techniques, with TSNE we have reduced the dimension and can see how each image is related to one another.\nIn the end, I have used a simple clustering technique called Kmeans to segregate the different data points and compare it with the original.\nThe related notebook is [here](https://www.kaggle.com/mrinath/image-analysis-with-tsne-and-kmeans?scriptVersionId=40804344)\n\nTo know more about TSNE you can watch [https://www.youtube.com/watch?v=NEaUSP4YerM](url)\n and you can find more on Kmeans [here.](https://www.youtube.com/watch?v=4b5d3muPQmA)Thank You!!"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "971555": "\"Let the neural network sleep\", recently I have come out with this statement in some article. The article said that the trend of just making an end to end neural network model without understanding the data is increasing creating many others and mysteries of the data unsolved.\nSo simply put, here I have tried to analyse the Image data provided. I have used a famous Dimension reducing algorithm called TSNE to make each image fit to a cartesian coordinate system, that is each image is represented as a point having x and y coordinate. \nAn image is a multidimensional data hence very difficult to understand and to use simple machine learning techniques, with TSNE we have reduced the dimension and can see how each image is related to one another.\nIn the end, I have used a simple clustering technique called Kmeans to segregate the different data points and compare it with the original.\nThe related notebook is [here](https://www.kaggle.com/mrinath/image-analysis-with-tsne-and-kmeans?scriptVersionId=40804344)\n\nTo know more about TSNE you can watch [https://www.youtube.com/watch?v=NEaUSP4YerM](url)\n and you can find more on Kmeans [here.](https://www.youtube.com/watch?v=4b5d3muPQmA)Thank You!!"
  }
}