Senseforth RND Engineer Interview Questions, Process, and Tips

Optimizers exist in FFNs to update weights, but many exist due to different optimization techniques and trade-offs.

• Different optimizers use different optimization techniques such as momentum, adaptive learning rates, and regularization.

• Optimizers have different trade-offs such as convergence speed, generalization, and robustness to noisy data.

• The choice of optimizer depends on the specific problem and data set.

• Examples...read more

Layer normalisation is a technique used to normalise the inputs to each layer of a feedforward neural network.

• It is similar to batch normalisation but normalises the inputs to each layer instead of the entire batch.

• It helps in reducing the internal covariate shift problem.

• It can be applied to any type of activation function.

• It is particularly useful in recurrent neural networks.

• Example: LayerNorm in PyTorch.

Build an airline system to recommend routes from city 1 to city 2 with direct and connecting flights.

• Create a graph with cities as nodes and connections as edges

• Use Dijkstra's algorithm to find shortest path between city 1 and city 2

• For connecting flights, find all possible paths with one or more stops

• Sort and recommend routes based on total distance and number of stops

This function takes a string input and returns all possible combinations of characters in that string without repeated letters.

• Use the itertools module to generate all possible permutations of the string.

• Filter out the permutations that have repeated letters using a set.

• Convert the filtered permutations into a list of strings.