Deep Learning and AI Superhero

Practical Exercises Chapter 6

Exercise 1: Implement a Simple RNN for Sequence Classification

Task: Implement a simple RNN to classify sequences of numbers. Use synthetic data where each sequence is classified as positive if the sum of the elements is above a threshold, and negative otherwise.

Solution:

import torch

import torch.nn as nn

import torch.optim as optim

from torch.utils.data import DataLoader, TensorDataset

# Generate synthetic data (binary classification based on sequence sum)

def generate_data(num_samples=1000, sequence_length=10, threshold=5):

X = torch.randint(0, 3, (num_samples, sequence_length)).float()

y = (X.sum(dim=1) > threshold).float()

return X, y

# Define the RNN model

class SimpleRNN(nn.Module):

def __init__(self, input_size ...