TensorFlow: Machine Learning for Everyone, Rajat Monga 20160222

Rajat Monga, TensorFlow Technical Lead & Manager, Google
TensorFlow™ is an open source software library for numerical computation using data flow graphs. The flexible architecture allows you to deploy computation to one or more CPUs or GPUs in a desktop, server, or mobile device with a single API. TensorFlow was developed by researchers and engineers working on the Google Brain Team for the purposes of conducting machine learning and deep neural networks research.
Source Code:

Speaker Bio
Rajat Monga works on the Google Brain Team, where he is the Technical Lead and Manager for TensorFlow – an open source machine learning library, and the center of Google’s efforts at scaling up deep learning. He is particularly interested in enabling smarter devices. Prior to Google, as the Chief Architect and Director of Engineering at Attributor, Rajat hired the founding engineering team, and led the labs and operations to design and build Attributor’s core matching engine. As a veteran developer Rajat has worked at eBay, Infosys, and a number of startups.


  1. auro tripathy says:

    If you want a working version at 20:36 …

    import tensorflow as tf
    with tf.Session(graph=tf.Graph()) as sess:
    examples = tf.constant([[1, 2, 3], [5, 6, 7]]) # 2X3 mat
    weights = tf.constant([[1, 1], [2, 2], [3, 3]]) # 3X2 mat
    biases = tf.constant([[1], [2]]) # 2X1 mat
    output = tf.matmul(examples, weights) + biases # 2X2 mat

  2. auro tripathy says:

    at 21:48 the code will run if ‘assign_op.run()’ is changed to
    ‘assign_p.eval()’. There’s a subtle difference between the ‘eval’ and ‘run’.

    import tensorflow as tf
    with tf.Session(graph=tf.Graph()) as sess:
    v = tf.Variable([[0.0, 0.0]]) # First init/assign
    print (v.eval()) # actual evaluation
    assign_op = v.assign(tf.constant([[1.0, 2.0]])) # reassign
    assign_op.eval() # new assignment happens
    print(v.eval()) # actual evaluation

  3. boreguarde says:

    Every sentence/paragraph starts off strong and clear, and then trails off
    to a mumble. Once he fades to a mumble, the last few words are

  4. Adam Zíka says:

    Great video. Thanks. I have a question. At time 24:49, why do call
    v.eval(). Isn’t assign_op.run() enough? Thank you.

  5. RandomUser20130101 says:

    Please focus the video recording on the slide content. Views of the speaker
    and the audience are not interesting at all.

Comments are closed.