Deep learning for biotechnology presentation

CS 886 Deep Learning for Biotechnology
Ming Li
Jan. 21, 2022

CONTENT
01. Word2Vec
02. Attention / Transformers
03. Pretraining: GPT / BERT
04. Deep learning applications in proteomics
05. Student presentations begin

03 LECTURE
THREE
Pretraining: GPT-2 and BERT

Avoiding Information bottleneck
03 GPT-2, BERT

Last time we introduced transformer
03 GPT-2, BERT

Transformers, GPT-2, and BERT
03
1. A transformer uses Encoder stack to model input, and uses Decoder stack
to model output (using input information from encoder side).
2. But if we do not have input, we just want to model the “next word”, we
can get rid of the Encoder side of a transformer and output “next word”
one by one. This gives us GPT.
3. If we are only interested in training a language model for the input for
some other tasks, then we do not need the Decoder of the transformer, that
gives us BERT.
GPT-2, BERT

GPT-2, BERT
03
1542M
762M
345M
117M parameters
GPT released June 2018
GPT-2 released Nov. 2019 with 1.5B parameters
GPT-3: 175B parameters trained on 45TB texts

GPT-2 in action
GPT-2, BERT
03
not
injure
injure
a
a
human
human
being
being

Byte Pair Encoding (BPE)
GPT-2, BERT
03
Word embedding sometimes is too high level, pure character embedding too
low level. For example, if we have learned
old older oldest
We might also wish the computer to infer
smart smarter smartest
But at the whole word level, this might not be so direct. Thus the idea is to
break the words up into pieces like er, est, and embed frequent fragments of
words.
GPT adapts this BPE scheme.

Byte Pair Encoding (BPE)
GPT-2, BERT
03
GPT uses BPE scheme. The subwords are calculated by:
1. Split word to sequence of characters (add </w> char)
2. Joining the highest frequency pattern.
3. Keep doing step 2, until it hits the pre-defined maximum number of sub-words
or iterations.
Example (5, 2, 6, 3 are number of occurrences)
{‘l o w </w>’: 5, ‘l o w e r </w>’: 2, ‘n e w e s t </w>’: 6, ‘w i d e s t </w>’: 3 }
{‘l o w </w>’: 5, ‘l o w e r </w>’: 2, ‘n e w es t </w>’: 6, ‘w i d es t </w>’: 3 }
{‘l o w </w>’: 5, ‘l o w e r </w>’: 2, ‘n e w est </w>’: 6, ‘w i d est </w>’: 3 } (est freq. 9)
{‘lo w </w>’: 5, ‘lo w e r </w>’: 2, ‘n e w est</w>’: 6, ‘w i d est</w>’: 3 } (lo freq 7)
…..

Masked Self-Attention (to compute more efficiently)
03 GPT-2, BERT

Masked Self-Attention
03 GPT-2, BERT
Note: encoder-decoder attention block is gone

Masked Self-Attention Calculation
03 GPT-2, BERT
Note: encoder-decoder attention block is gone
Re-use previous computation results: at any step, only
need to results of q, k , v related to the new output
word, no need to re-compute the others. Additional
computation is linear, instead of quadratic.

GPT-2 fully connected network has two layers (Example for GPT-2
small)
03 GPT-2, BERT
768 is small model size

GPT-2 has a parameter top-k, so that we sample words from top k
(highest probability from softmax) words for each output
03 GPT-2, BERT

This top-k parameter, if k=1, we would have output like:
03 GPT-2, BERT
The first time I saw the new version of the game, I was
so excited. I was so excited to see the new version of
the game, I was so excited to see the new version of

GPT Training
03 GPT-2, BERT
GPT-2 uses unsupervised learning approach
to training the language model.
There is no custom training for GPT-2, no
separation of pre-training and fine-tuning
like BERT.

A story generated by GPT-2
03 GPT-2, BERT
“The scientist named the population, after their distinctive horn, Ovid’s Unicorn. These four-
horned, silver-white unicorns were previously unknown to science.
Now, after almost two centuries, the mystery of what sparked this odd phenomenon is finally
solved.
Dr. Jorge Pérez, an evolutionary biologist from the University of La Paz, and several companions,
were exploring the Andes Mountains when they found a small valley, with no other animals or
humans. Pérez noticed that the valley had what appeared to be a natural fountain, surrounded
by two peaks of rock and silver snow.
Pérez and the others then ventured further into the valley. `By the time we reached the top of
one peak, the water looked blue, with some crystals on top,’ said Pérez.
Pérez and his friends were astonished to see the unicorn herd. These creatures could be seen
from the air without having to move too much to see them – they were so close they could
touch their horns."

Transformer / GPT prediction
03 GPT-2, BERT

GPT-2 Application: Translation
03 GPT-2, BERT

GPT-2 Application: Summarization
03 GPT-2, BERT

Using wikipedia data
03 GPT-2, BERT

BERT (Bidirectional Encoder Representation from Transformers)
03 GPT-2, BERT

Model input dimension 512
Input and output vector size
03 GPT-2, BERT

BERT pretraining
03 GPT-2, BERT
ULM-FiT (2018): Pre-training ideas, transfer learning in NLP.
ELMo: Bidirectional training (LSTM)
Transformer: Although used things from left, but still missing
from the right.
GPT: Use Transformer Decoder half.
BERT: Switches from Decoder to Encoder, so that it can use
both sides in training and invented corresponding training
tasks: masked language model

BERT Pretraining Task 1: masked words
03 GPT-2, BERT
Out of this 15%,
80% are [Mask],
10% random words
10% original words

BERT Pretraining Task 2: two sentences
03 GPT-2, BERT

BERT Pretraining Task 2: two sentences
03 GPT-2, BERT
50% true second sentences
50% random second sentences

Fine-tuning BERT for other specific tasks
03 GPT-2, BERT
SST (Stanford
sentiment treebank):
215k phrases with
fine-grained
sentiment labels in
the parse trees of
11k sentences.
MNLI
QQP (Quaro Question Pairs)
Semantic equivalence)
QNLI (NL inference dataset)
STS-B (texture similarity)
MRPC (paraphrase, Microsoft)
RTE (textual entailment)
SWAG (commonsense inference)
SST-2 (sentiment)
CoLA (linguistic acceptability
SQuAD (question and answer)

NLP Tasks: Multi-Genre Natural Lang. Inference
03 GPT-2, BERT
MNLI: 433k
pairs of
examples,
labeled by
entailment,
neutral or
contraction

NLP Tasks (SQuAD -- Stanford Question Answering Dataset):
03 GPT-2, BERT
Sample: Super Bowl 50 was an American football game to
determine the champion of the National Football League (NFL)
for the 2015 season. The American Football Conference (AFC)
champion Denver Broncos defeated the National Football
Conference (NFC) champion Carolina Panthers 24–10 to earn
their third Super Bowl title. The game was played on February 7,
2016, at Levi's Stadium in the San Francisco Bay Area at Santa
Clara, California. As this was the 50th Super Bowl, the league
emphasized the "golden anniversary" with various gold-themed
initiatives, as well as temporarily suspending the tradition of
naming each Super Bowl game with Roman numerals (under
which the game would have been known as "Super Bowl L"), so
that the logo could prominently feature the Arabic numerals 50.
Which NFL team represented the
AFC at Super Bowl 50?
Ground Truth Answers: Denver
Broncos
Which NFL team represented the
NFC at Super Bowl 50?
Ground Truth Answers: Carolina
Panthers

Add indices for sentences and paragraphs
SegaTron/SegaBERT
H. Bai, S. Peng, J. Lin, L. Tan, K. Xiong, W. Gao, M. Li: SgaTron: Segment-aware transformer for language modeling
and understanding. AAAI’2021

Conversion speed much faster：

Testing on GLUE dataset
H. Bai, S. Peng, J. Lin, L. Tan, K. Xiong, W. Gao, M. Li: SgaTron: Segment-aware transformer for language modeling
and understanding. AAAI’2021

Reading comprehesion – SQUAD tasks
F1 = 2 (P*R) / (P+R), P is precision, R is recall, all in percentage, EM – exact match

Feature Extraction
03 GPT-2, BERT
We start with
independent
word embedding
at first level
We end up with
some embedding
for each word
related to current
input

Feature Extraction, which embedding to use?
03 GPT-2, BERT

What we have learned
03 GPT-2, BERT
1. Model size matters (345 million parameters is better than 110
million parameters).
2. With enough training data, more training steps imply higher
accuracy
3. Key innovation: Learning from unannotated data.
4. In biotechnology, we also have a lot of such data (for example
meta-genomes).

Literature & Resources for Transformers
03
Resources:
OpenAI GPT-2 implementation: https://github.com/openai/gpt-2
BERT paper: J. Devlin et al, BERT, pretraining of deep bidirectional
transformers for language understanding. Oct. 2018.
ELMo paper: M. Peters, et al, Deep contextualized word representation, 2018
ULM-FiT paper: Universal language model fine-tuning for text classification. J.
Howeard, S. Ruder., 2018
Jay Alammar, The illustrated GPT-2, https://jalammar.github.io/illustrated-
gpt2/
GPT-2, BERT

Deep learning for biotechnology presentation

More Related Content

Similar to Deep learning for biotechnology presentation (20)

Recently uploaded (20)

Deep learning for biotechnology presentation

Editor's Notes