What are the current employment prospects for artificial intelligence? Is it easy to find a job after completing courses related to artificial intelligence? At present, the domestic artificial intelligence field is developing rapidly, and major enterprises are actively deploying artificial intelligence technology applications. If you want to engage in artificial intelligence-related work, you must not only understand the overall development direction of the country. The only stepping stone to entering a company is the foundation. Today, I found some interview questions about artificial intelligence in the Dark Horse Programmer Forum. I hope to be of some help to friends who are seeking knowledge in the near future.
Artificial intelligence interview questions sharing (including answers)
1. What are the four commonly used cross entropies in the deep learning framework TensorFlow?
Answer: tf.nn.weighted_cross_entropy_with_logits
tf.nn.sigmoid_cross_entropy_with_logits
tf.nn.softmax_cross_entropy_with_logits
tf.nn.sparse_softmax_cross_entropy_with_logits
2. What is over-fitting and what measures are there to avoid over-fitting?
Answer: Overfitting: In machine learning, when we test the model, we improve the expressiveness on the training data set,
but the expressiveness on the training set decreases.
Solutions:
1. Regularization;
2. Adjust the parameters in the process of training the model. The learning rate should not be too high;
3. Cross-validate the data;
4. Select the percentage of the test set data that is suitable for the training set, and select the appropriate stop training criteria to make the machine training appropriate;
5. In the neural network model, we The weight can be reduced;
3. What is a kernel function?
The kernel function infers the linear inseparable features to the high-level feature space, so that the support vector machine is linearly separable in this high-dimensional space, that is, the kernel function can be used to map to the high-dimensional space and solve the nonlinear classification problem. Including linear kernel function, polynomial kernel function, Gaussian kernel function, etc., among which Gaussian kernel function is the most commonly used.
4. What are the common kernel functions in the deep learning framework TensorFlow?
The SVM Gaussian kernel function should be that if you want to divide the nonlinear data set and change the linear classifier to the data set, you must change the kernel function in the SVM loss function.
linear kernel function
polynomial kernel function
5. What are the advantages of the Naive Bayes method?
Naive Bayes has a stable classification efficiency
. performs well for small-scale data and can handle multi-classification problems. When the data exceeds the memory, incremental training
is not sensitive to missing data. The algorithm is relatively simple and is often used for text classification. .
6. What is the standard method of supervised learning?
All regression algorithms and classification algorithms belong to supervised learning and clearly give initial values that have features and labels in the training set, and obtain a model through training, which can make predictions when facing data with only features but no labels.
7. What is the choice of model in machine learning?
According to the performance of a set of models of different complexity, the best model is selected from a certain model. After selecting a best model, evaluate its prediction errors and other evaluation indicators on the new data.
8. What are the advantages and disadvantages of the graph database Neo4J?
Advantages: 1. Faster database operations, provided that the amount of data is large enough.
2. The data is more intuitive, and the corresponding SQL statements are easier to write.
3. More flexible, no matter what new data needs to be stored, it is the same node, only the node attributes and edge attributes need to be considered.
4. The operation of the database will not decrease significantly as the database grows.
Disadvantages: 1. Very slow insertion speed.
2. Oversized nodes. When a node has a lot of edges,
the operation speed of the node will be greatly reduced
9. What is the connection and difference between LR and SVM?
Both are classification algorithms
If the kernel function is not considered, LR and SVM are both linear classification algorithms, which means that their classification decision surfaces are linear.
Both LR and SVM are supervised learning algorithms
. The loss functions of LR and SVM are different.
SVM only considers the points near the local boundary, while LR considers the global. Points far away also play a role in determining the boundary.
10. What is clustering and its application scenarios?
Clustering refers to dividing a thing into several parts that do not overlap each other according to certain criteria. In machine learning, clustering refers to a standard. This criterion is usually similarity. The sample is divided into several parts. Those with a high degree of similarity are gathered together, and those with a low degree of similarity are separated from each other.
Clustering application scenarios, complete job search information (about 100,000 high-quality resumes, some of which contain complete fields, and some of the resumes are left blank in fields such as academic qualifications, company size, salary, etc. Hope to learn and code the data And testing, dig out the trend and law of the job path, form an algorithm model, and predict the information that is blank in the data.)
The above are 10 interview questions about artificial intelligence. The follow-up will share with you more about artificial intelligence learning, interviews and other aspects.
I will continue to share some interview tips with you later, remember to give me a thumbs up!