Does clustering actually reduce the number of rows in a dataset?
I am reading the book “grokking Machine Learning” by Luis G. Serrano and came across the following sentence:
Machine Learning Confusion Matrix Explanation of Matrix
`The following shows a few Confusion Matrices. The TP, FP, FN & TN are marked (using ChatGPT)
Wrong machine learning predictions at every run
I trained a machine learning model, and a few of the same rows (around 100 out of 1800) in my test set are giving wrong predictions at every run (10 times) with different seeds. Is there anything I should do about it e.g. throw it in the train set in exchange for some data from the train set to test set or leave it as it is?
Would passing correctly predicted images by a model back to the model for training purposess yield anything useful to us?
I do not have the required technical knowledge to try and test the metrics of a model to compare results and hypothesize, hence I am relying on the wisdom of the community.
I tried to look for answers however did not find anything useful (perhaps I phrased my question incorrectly).
Any guidance on the above topic would be appreciated.
In a generative model, how to factorize the likelihood of observing the data using the DAG dependencies?
In Machine Learning, considering a generative model represented by the following Direct Acyclc Graph (DAG):
increasing the input dataset and finding the corresponding increased output by training the current input and outputs
The input excel file(trainn.xlsx) represents the car percentage and the output excel file(pred.xlsx) represents the different output values from input values.There are 160 inputs and 160 outputs.I want to generate more 340 inputs and corresponding 340 outputs(for every column) by training the current dataset.Is it possible to increase the values to 500 for both input and output by generating new datas??My current code:
Classification on machine learning algorithms
I have been studying machine learning for past three months and came across the statement “Machine learning models can be classified as generative or descriptive, probabilistic vs non-probabilistic and parametric vs non-parametric.”
Getting high loss on the model that I created
I can share the full data on the stack, so I would just share the code below.
I have been trying for the past few days on this data but it seems that the loss is coming around 18000 or 15000 which is to high. My colleagues told me to use neural network for this data.
Is machine learning good for predicting values like production speed?
I wonder if machine learning would be a good way to predict production performance meters (like machine speed pcs/h, machine setup time, waste). The product is carton boxes, so quite simple thing from construction point (easy to classify)
The scenario is that there are several types of machines, different designs of boxes (which can be grouped into lets say 10 types), so those parameters are not numeric, then we have numeric parameters like carton grammage, box width,length, height, number of colors. So I am looking to train a model which after providing machine type, box type, grammage, width,length etc would predict machine setup time, runnng speed i pcs/h, waste %.
If machine learing is a good tool for that purpose then please can you tell me which kind of models would suit the best.
Are the predictions made by LLMs based on probability?
I’m studying machine learning and will like to understand how LLMs make their predictions. Are these predictions probabilistic or they are absolute truth.