zhenye-na / advanced-deep-learning-and-reinforcement-learning-deepmind

🎮 Advanced Deep Learning and Reinforcement Learning at UCL & DeepMind | YouTube videos 👉

Hello there 👋

#!/usr/bin/python
# -*- coding: utf-8 -*-


class SoftwareEngineer:

    def __init__(self):
        self.name = "Zhenye Na"
        self.role = "Software Engineer"
        self.language_spoken = ["zh_CN", "en_US"]

    def say_hi(self):
        print("Thanks for dropping by, hope you find some of my work interesting.")


me = SoftwareEngineer()
me.say_hi()

📝 Blogs

• Personal website and blog: https://zhenye-na.github.io/
• Notes of AWS Solution Architecture - Associate: https://zhenye-na.github.io/aws-certs-cheatsheet/
• [WIP] LeetCode Solutions & Notes: https://zhenye-na.github.io/leetcode/

📔 Latest Blog posts

• Java Concurrent Programming 1 - Fundamentals
• Introduction to Dependency Injection in Java
• Distributed Transactions and Sagas in Microservices
• Design Patterns: Strategy Pattern
• Java Best Practices - Logging

🔧 Technologies & Tools

Cloud Services:

Programming Languages:

Tools and Services:

🏆 GitHub Trophies

🗂️ Highlight Projects

advanced-deep-learning-and-reinforcement-learning-deepmind's Issues

DL Homework 3

In Q1 you write that exactly one possible scenario exists: The outputs of the input and forget gates must be 0 or 1. Since the gates use a sigmoid activation function it is numerically impossible to reach exactly the values 0 and 1. It is only possible to get close to the values, but in every such scenario, neither only the input nor a previous state can be taken into account only.

Furthermore I don't see that you have implemented GRU. You only use BasicLSTMCell in the whole task.