{"id":199,"date":"2024-12-08T20:49:22","date_gmt":"2024-12-08T20:49:22","guid":{"rendered":"https:\/\/aulendil.net\/hallucinations\/?p=199"},"modified":"2024-12-08T20:49:22","modified_gmt":"2024-12-08T20:49:22","slug":"wrong-again","status":"publish","type":"post","link":"https:\/\/aulendil.net\/hallucinations\/wrong-again\/","title":{"rendered":"Wrong again"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

Once upon a time, I was doing some analysis<\/a> of a random forest. At least I thought it was about the random forest algorithm\u2026 I was so wrong. What I was trying to understand and what didn\u2019t feel right was not a random forest; it was the Random Subspace Method. It\u2019s still a forest of decision trees, but it works quite a little differently. Random forests bootstrap features at each node to generate the next decision node, while random subspace method bootstraps features once per tree\u2026 so it\u2019s something quite different.<\/p>\n\n\n\n

That\u2019s one lesson\u2014a small mistake in reading the documentation, in listening to the lecture. Exactly one word misunderstood\u2026 and I was on the wrong path, examining something very different. I just thought that random forest picks a subset of the features for the whole tree, not for every node. That reminds me of so many problems at work, where misreading one word of a hundred-page-long documentation leads to some not very much desired results\u2026<\/p>\n\n\n\n

Now\u2026 my insights were still interesting. Something I called normal bias was what I thought I discovered through this journey. But it was just a phantom that came from another small mistake. A bug in the code <\/a>gave me such wrong results. One line of code, one variable not assigned correctly. And ironically, everything seemed to work; it even made sense, a lot of sense. It was intuitive, reasonable\u2026 After I discovered this bug and fixed it, my results, more aligned with the real value of those algorithms, are more puzzling than results produced with the bug. Randomly unequal distribution of data with just random bias, where some features are more represented just because of sheer chance, gives better results time and time again than predictors with features distribution adjusted for random bias. It makes no sense, but it is so\u2026 There is no point in fighting with the facts.<\/p>\n\n\n\n

One would say, I was wrong – end of the road, time to move on. But it doesn’t feel right for me\u2026 It’s just another question: ‘Why’? Why is it so weird? It feels just wrong to abandon those questions. They are there, screaming for the answer. Does this ever end?<\/p>\n\n\n\n

Initially, I planned to abandon this little research project I started during my training. I have so many other plans. But now I wonder. There are so many questions, so many projects to do. But here is the question that asks for an answer. And questions are nasty little things\u2014they take my sleep away, they don\u2019t let me rest. <\/p>\n\n\n\n

It\u2019s easy to write stuff on paper or in the mind. They work, they are coherent, they make sense, nothing contradicts it\u2026 but once data comes, some validation is done\u2026 stuff happens. Nasty little questions that won’t let me walk away – they steal my peace.<\/p>\n","protected":false},"excerpt":{"rendered":"

Once upon a time, I was doing some analysis of a random forest. At least I thought it was about the random forest algorithm\u2026 I was so wrong. What I was trying to understand and what didn\u2019t feel right was not a random forest; it was the Random Subspace Method. It\u2019s still a forest of […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"_links":{"self":[{"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/posts\/199"}],"collection":[{"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/comments?post=199"}],"version-history":[{"count":1,"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/posts\/199\/revisions"}],"predecessor-version":[{"id":201,"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/posts\/199\/revisions\/201"}],"wp:attachment":[{"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/media?parent=199"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/categories?post=199"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aulendil.net\/hallucinations\/wp-json\/wp\/v2\/tags?post=199"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}