Monday, July 25th, 2022 - Saturday, July 30th, 2022

In paper [1], the accuracy of autistic children prediction is higher than 90%. In addition to the amazing result, the dataset used in this paper also shocked me. The author uses the Autism_Image_Data from Kaggle and the paper was published in Computational and Mathematical Methods in Medicine, a magazine of the Q2 section.

Isn’t making a prediction by using a dataset from Kaggle too simple to be recorded in a paper published in the Q2 section?

Anyway, it was a good chance for practicing. There are some demos available on Kaggle but they all use tensorflow.keras.preprocessing which is deprecated. It cost me about a week to recode using tensorflow.keras.dataset.From my experiment experience, the tensorflow.keras.dataset seems to be about 10% faster than the tensorflow.keras.preprocessing.

In my experiment, I could only achieve about 75% of accuracy. Fine. I just skipped it since it was just a test.

During the exam, the Visual Studio Code always popped out an error saying the kernel died. I read the jupyter log line by line but didn’t find the solution. Finally, I used pip to install jupyter and saw the exact error in the browser instead of the Visual Studio Code. The error said I missed the zlibwapi.dll. Okay, the solution is on its way.

It had been a long time since I didn’t code. The error made me worry about whether my laptop had some problems.

[1] Z. A. T. Ahmed et al., “Facial Features Detection System To Identify Children With Autism Spectrum Disorder: Deep Learning Models.,” Comput Math Methods Med, vol. 2022, p. 3941049, 2022, doi: 10.1155/2022/3941049.