rish-16

CS2030S Week 3 Lecture

Notes from Week 3 Lecture 1 on Object Oriented Programming

java.lang.Object class

• Mother of all objects and classes
• All classes implicitly inherit from the Object class

Polymorphism

Literally means "many forms" – allows us to decide and change already existing behaviour. At runtime, when methods are used for the polymorphed class, it'll call the methods that are specific to said object.

rish-16

CS2040S Week 2 Lecture 2

Notes from Week 2 Lecture 2 on Data Structures and Algorithms.

Problem Solving

Peak Finding

Aim is to find the global maximum and ignore local maxima.

inputs: Array A[0 ... n-1]
output: Maximum element in A

rish-16

CS2040S Week 2 Lecture 1

Notes from Week 2 on Data Structures and Algorithms.

Miscellaneous

• Midterm Exam: Week 7

Algorithm Analysis

Time Complexity Analysis

As we scale, we want to see which algorithms are faster for large inputs ie. asymptotic performance. T(n) is the running time of an algorithm.

rish-16

CS2030S Week 2

Abstraction Barrier

Something that stands between the client/user and implementation. The implementer declares the variables and methods and the user uses this implementation object via abstraction (only public things can be accessed). This protects the internal state attributes from the user.

Information Hiding

Information Hiding enables implementers to make changes within the class without affecting the user much. The user continues to use the class object the same way but the internal implementation might have changed.

In Python, there are no access control rules. However, Java provides mechanisms to the compiler to enforce access control. We can add modifiers to variables and methods:

rish-16

CS2040S Week 1 Lecture 2

Notes from Week 1 on Data Structures and Algorithms.

Not covered

• Lambda expressions
• Type inference
• Packages

Basics of OOP

Programming paradigms

rish-16

Deploying on repl.it

1. Visit repl.it
2. Click on any of the 3 servers, rhapp_facilities, rhapp_events, or rhapp_laundry
3. Add changes to document
4. Click the Stop button at the top
5. Click the Run button at the top – a browser-like window should open with the default endpoint loaded
6. Test installation by pinging the endpoints with Postman (your repl.it window should show 200 codes in the console)

Important Notes

rish-16

# Select appropriate distribution strategy
if tpu:
  tf.tpu.experimental.initialize_tpu_system(tpu)
  strategy = tf.distribute.experimental.TPUStrategy(tpu, steps_per_run=128)
  print('Running on TPU ', tpu.cluster_spec().as_dict()['worker'])  
else:
  strategy = tf.distribute.get_strategy() # Default strategy that works on CPU and single GPU
  print('Running on CPU instead')
print("Number of accelerators: ", strategy.num_replicas_in_sync)

rish-16

# Detect hardware
try:
  tpu = tf.distribute.cluster_resolver.TPUClusterResolver() # TPU detection
except ValueError: # If TPU not found
  tpu = None

rish-16

history = model.fit(x_train, 
		    y_train,
                    epochs=20,
                    steps_per_epoch=50
                    )

model.save_weights('./mnist_model.h5', overwrite=True)

rish-16

with strategy.scope():
  """
  This essentailly takes our model and makes it 
  compatible to train on a TPU.
  """
  model = tf.keras.models.Sequential()
  model.add(tf.keras.layers.Dense(512, input_size=[784,], activation='relu'))
  model.add(tf.keras.layers.Dense(256, activation='relu'))
  model.add(tf.keras.layers.Dense(128, activation='relu'))
  model.add(tf.keras.layers.Dense(64, activation='relu'))