[Pattern recognition] The learning problem

시작하기전에

본 포스팅은 패턴인식 수업 수강 후 Learning problem에 대한 기본적인 지식들에 대해 복습할 기회 제공을 위해 개인적으로 만든 복습 문제 및 정답 포스팅입니다.

Questions

1. What is the hypothesis? and what is the ‘g’ and ‘f’ in the learning problem usually.

2. What is the target function ‘f’?

3. What is the PLA?

4. How perceptron learn about data?

5. What is the most studied and most utilized type of learning?

6. How we possibly learn anything from mere inputs?

7. What is the reinforcement learning?

8. What is the data mining?

9. What is the difference between the data mining and machine learning?

10. What is the difference between statistical learning vs machine learning

Answer

1. The formula which can make the algorithm chooses the best linear fit to Data.

2. The formula which can classify the problem we want correctly.

3. Perceptron Learning Algorithm

4. Pick a incorrect point from $(x_n,y_n)$

5. Supervised learning (Training the hypothesis from the labeled data)

6. Clustering, hidden markov models, feature extraction(PCA, SVD and so on…)
   • We can estimate the upper bound when we know about the expectation of input.

7. The network including Input – some output – reward process for this output

8. A practical field that focuses on finding patterns, correlations, anomalies.
   • Ex) look at medical records to detect a long-term drug effect
9. Technically, the same
   • DM more emphasis on data analysis
   • DM has more huge DB
10. ML: subfield of articial intelligence, SL: subfield of statistics
    • ML: applications, prediction accuracy, SL: models and interpretability