Interview with 1MG.com DATA SCIENCE profile
In total, I gave 4 rounds of Interview for 1MG.
Round 1: Telephonic Round
Intermediate level interview round where I was asked about Naive Bayes, Bayes theorem, my projects, and Convolution Neural Network working in brief.
Round 2: Technical round 1 onsite
It was for more than 1 hour where I was asked about DCGANs, loss functions, Logistic regression in-depth knowledge, working of Naive-Bayes Classifier, Bayes theorem, and Total Probability theorem. Loss functions Mean square error, Log-Loss, Mean square log error were asked. Few more topics/algorithms were asked to know my knowledge in the field.
Question: For a breast cancer dataset, I applied three algorithms. One decision tree where I get 97% accuracy on the test set, Two random forest: 98% accuracy, Three Naive Bayes: 92% accuracy. Which algorithm is best among these?
Answer: None of these, Accuracy those not tell which algorithm is best. Precision and Recall is also important.
The algorithm may be correctly predicting that a person does not have cancer and test results are also -ve but this does not mean it will predict +ve test when a person really has cancer.
This case is known as accuracy Paradox.
Round 3: Technical round 2 onsite
Here, I was asked about application part where can Machine Learning be used in their platform. This was my practical and application knowledge tester. I had already gone through their website and app thoroughly so I knew where recommendation part can be used where Association rule methods can be used. Here we discussed apriori and Fp-growth tree algorithm.
Round 4: HR ROUND
After around 20mins, HR came up and asked me some questions about me. Went through my resume and within 2days I got selection msg.