程序代写代做代考 Bayesian AI ada deep learning game html COMP9444

COMP9444
Neural Networks and Deep Learning
What is a Neural Network?
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
2
COMP9444 20T3 Neuroanatomy
3
1b. Neuroanatomy
􏰀 massivelyparalleldistributedprocessormadeupofsimpleprocessing units
Why Neural Networks?
Sub-Symbolic Processing
􏰀 biologically inspired
􏰀 good learning properties
􏰀 continuous, nonlinear
􏰀 well adapted to certain tasks 􏰀 fault tolerant
􏰀 graceful degradation
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
Textbook, Section 9.10
COMP9444 20T3 Neuroanatomy 1
􏰀 knowledge acquired from environment through a learning process 􏰀 knowledge stored in the form of synaptic weights

COMP9444 20T3 Neuroanatomy
4
COMP9444 20T3 Neuroanatomy 5
Theories about Intelligence
Artificial Intelligence Origins
􏰀 380BC Plato (Rationalism – innateness)
􏰀 330BC Aristotle (Empricism – experience) 􏰀 1641 Descartes (mind-body Dualism)
􏰀 1781 Kant (Critique of Pure Reason)
􏰀 1899 Sigmund Freud (Psychology)
􏰀 1953 B.F. Skinner (Behaviourism)
􏰀 1642 Blaise Pascal (mechanical adding machine)
􏰀 1694 Gottfried Leibniz (mechanical calculator)
􏰀 1769 Wolfgang von Kempelen (Mechanical Turk)
􏰀 1837 Charles Babbage & Ada Lovelace (Difference Engine) 􏰀 1848 George Boole (the Calculus of Logic)
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
COMP9444 20T3 Neuroanatomy
6
COMP9444 20T3
Neuroanatomy
7
Neural Network Origins
Serial Symbolic AI
􏰀 1943 McCulloch & Pitts (neuron models) 􏰀 1948 Norbert Wiener (Cybernetics)
􏰀 1948 Alan Turing (B-Type Networks)
􏰀 1955 Oliver Selfridge (Pattern Recognition) 􏰀 1962 Hubel and Wiesel (visual cortex)
􏰀 1956 Newell & Simon (Logic Theorist) 􏰀 1959 John McCarthy (Lisp)
􏰀 1959 Arther Samuel (Checkers)
􏰀 1965 Joseph Weizenbaum (ELIZA)
􏰀 1962 Frank Rosenblatt (Perceptron)
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
􏰀 1879 Gottlob Frege (Predicate Logic) 􏰀 1950 Turing Test
􏰀 1956 Dartmouth conference
􏰀 1967 Edward Feigenbaum (Dendral)

COMP9444 20T3 Neuroanatomy
8
COMP9444 20T3 Neuroanatomy 9
Neural Network “Dark Ages”
Knowledge-Based Systems
􏰀 1969 Minsky & Papert published Perceptrons, emphasizing the limitations of neural models, and lobbied agencies to cease funding neural network research.
􏰀 1970s and early 1980s, AI research focused on symbolic processing, Expert Systems
􏰀 from 1969 to 1985 there was very little work in neural networks or machine learning.
◮ ◮
combinatorial explosion in search spaces
􏰀 a few exceptions, e.g. Stephen Grossberg, Teuvo Kohonen (SOM), Paul Werbos.
difficulty of formalising everyday knowledge as well as expert knowledge
COMP9444 ⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
COMP9444 20T3 Neuroanatomy
10
COMP9444 20T3
Neuroanatomy
11
Neural Network Renaissance
Applications of Deep Learning
􏰀 1986 Rumelhart, Hinton & Williams (multi-layer, backprop) 􏰀 1989 Dean Pomerleau (ALVINN)
􏰀 late 1980’s renewed enthusiasm, hype
􏰀 1990s more principled approaches
􏰀 Image processing ◮ classification ◮ segmentation
􏰀 2000’s SVM, Bayesian models became more popular 􏰀 2010’s deep learning networks, GPU’s
􏰀 2020’s spiking networks(?)
􏰀 Combining images and tex ◮ automatic captioning
COMP9444
⃝c Alan Blair, 2020
◮ Deep Q-Learning COMP9444
⃝c Alan Blair, 2020
􏰀 Some commercial success, but ran into difficulties:
􏰀 Language processing ◮ translation
◮ semantic disambiguation ◮ sentiment analysis
􏰀 Game playing ◮ AlphaGo

COMP9444 20T3 Neuroanatomy 12
COMP9444 20T3
Neuroanatomy
13
History of Deep Learning
Neuroanatomy
Two perspectives on the history of Deep Learning Viewpoint 1: Focusing on recent work (after 2012)
􏰀 Central Nervous System ◮ Brain
https://www.cs.toronto.edu/~hinton/absps/NatureDeepReview.pdf
◮ Spinal cord
􏰀 Peripheral Nervous System
Viewpoint 2: Focusing on earlier work (before 2012)
◮ ◮ ◮
Somatic nervous system Autonomic nervous system Enteric nervous system
http://people.idsia.ch/~juergen/deep-learning-overview.html
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
14
COMP9444 20T3
Neuroanatomy
15
Brain Regions
Cerebral Cortex
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020

COMP9444 20T3 Neuroanatomy 16
COMP9444 20T3 Neuroanatomy 17
Cerebral Cortex
Brain Stem
COMP9444 20T3
Neuroanatomy 18
COMP9444 20T3 Neuroanatomy 19
􏰀 “cortex” from Latin word for “bark” (of tree)
􏰀 general term for area of brain between the thalamus and spinal cord 􏰀 includes medulla, pons, tectum, reticular formation and tegmentum 􏰀 functions: breathing, heart rate, blood pressure, and others
􏰀 cortexisasheetoftissuemakingupouterlayersofbrain,2-6cmthick
􏰀 right and left sides connected by corpus callosum
􏰀 functions: thought, voluntary movement, language, reasoning, perception
Cerebellum
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
Midbrain
Hypothalamus
􏰀 functions: vision, audition, eye movement, body movement Thalamus
􏰀 composed of several different areas at the base of the brain
􏰀 receives sensory information and relays it to the cerebral cortex
􏰀 also relays information from the cerebral cortex to other areas of the
brain, and the spinal cord
􏰀 functions: sensory integration, motor integration
COMP9444 ⃝c Alan Blair, 2020
􏰀 from Latin word for “little brain”
􏰀 functions: movement, balance, posture
􏰀 the size of a pea (about 1/300 of the total brain weight)
􏰀 functions: body temperature, emotions, hunger, thirst, circadian rhythms
COMP9444
⃝c Alan Blair, 2020

COMP9444 20T3
Neuroanatomy
20
COMP9444 20T3 Neuroanatomy 21
Limbic System
Limbic System
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
22
COMP9444 20T3 Neuroanatomy 23
Brain Functions
Neurons as Body Cells
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
􏰀 group of structures including amygdala, hippocampus, mammillary bodies and cingulate gyrus
􏰀 important for controlling the emotional response to a given situation
􏰀 hippocampus also important for memory
􏰀 functions: emotional behaviour
􏰀 The body is made up of billions of cells. Cells of the nervous system, called neurons, are specialized to carry “messages” through an electrochemical process.
􏰀 The human brain has about 100 billion neurons, and a similar number of support cells called “glia”.
􏰀 Neurons are similar to other cells in the body in some ways, such as: ◮ neurons are surrounded by a cell membrane
◮ neurons have a nucleus that contains genes (DNA)
◮ neurons carry out basic cellular processes like protein synthesis and energy production

COMP9444 20T3 Neuroanatomy
24
COMP9444 20T3 Neuroanatomy 25
Structure of a Typical Neuron
Neurons versus Body Cells
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
26
COMP9444 20T3 Neuroanatomy 27
Variety of Neuron Types
Axons and Dendrites
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
􏰀 Neurons have specialized extensions called dendrites and axons Dendrites bring information to the cell body, while axons take information away from the cell body.
􏰀 Theaxonofoneneuroncanconnecttothedendriteofanotherneuron through an electrochemical junction called a synapse.
􏰀 Most neurons have only one axon, but the number of dendrites can vary widely:
◮ Unipolar and Bipolar neurons have only one dendrite ◮ Purkinje neurons can have up to 100,000 dendrites
􏰀 Dendrites are typically less than a millimetre in length
􏰀 Axons can vary in length from less than a millimetre to more than a
metre (motor neurons)
􏰀 Long axons are sometimes surrounded by a myelinated sheath, which prevents the electrical signal from dispersing, and allows it to travel faster (up to 100 m/s).

COMP9444 20T3
Neuroanatomy
28
COMP9444 20T3 Neuroanatomy 29
Synapse
Synapses and Ion Channels
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
30
COMP9444 20T3 Neuroanatomy 31
Ion Channel
The Big Picture
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
􏰀 electrical pulse reaches the endbulb and causes the release of neurotransmitter molecules from little packets (vesicles) through the synaptic membrane
􏰀 transmitter then diffuses through the synaptic cleft to the other side
􏰀 when the neurotransmitter reaches the post-synaptic membrane, it
causes a change in polarisation of the membrane
􏰀 the change in potential can be excitatiory (moving the potential towards the threshold) or inhibitory (moving it away from the threshold)
􏰀 human brain has 100 billion neurons with an average of 10,000 synapses each
􏰀 latency is about 3-6 milliseconds
􏰀 therefore, at most a few hundred “steps” in any mental computation,
but massively parallel

COMP9444 20T3 Neuroanatomy 32
COMP9444 20T3 Neuroanatomy 33
Hubel and Weisel – Visual Cortex
Convolutional Networks
􏰀 cells in the visual cortex respond to lines at different angles
􏰀 cells in V2 respond to more sophisticated visual features
􏰀 Convolutional Neural Networks are inspired by this neuroanatomy 􏰀 CNN’s can now be simulated with massive parallelism, using GPU’s
If we can identify features such as feather, eye, or beak which provide useful information in one part of the image, then those features are likely to also be relevant in another part of the image.
COMP9444
⃝c Alan Blair, 2020
COMP9444 ⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
34
COMP9444 20T3 Neuroanatomy 35
Convolutional Filters
Recurrent Neural Networks
COMP9444
⃝c Alan Blair, 2020
First Layer
Second Layer
Third Layer
􏰀 useful for processing language or other temporal sequences
COMP9444 ⃝c Alan Blair, 2020
Suppose we want to classify an image as a bird, sunset, dog, cat, etc.
We can exploit this regularity by using a convolution layer which applies the same weights to different parts of the image.
􏰀 can “unroll” a recurrent architecture into an equivalent feedforward architecture, with shared weights

COMP9444 20T3
Neuroanatomy
36
COMP9444 20T3 Neuroanatomy 37
Deep Q-Learning
Autoencoder Networks
COMP9444
⃝c Alan Blair, 2020
forced to compress the data in some way COMP9444
⃝c Alan Blair, 2020
COMP9444 20T3
Neuroanatomy
38
COMP9444 20T3
Neuroanatomy
39
Generative Adversarial Networks
Spiking Neurons
COMP9444
⃝c Alan Blair, 2020
COMP9444
⃝c Alan Blair, 2020
􏰀 output is trained to reproduce the input as closely as possible
􏰀 activations normally pass through a bottleneck, so the network is
􏰀 biological neurons spike in different patterns
(quiescent, persistent, sporadic)
􏰀 spike timing might carry important information
􏰀 most NN models ignore timing information, but some work has been done on spiking network models
􏰀 in the future, special hardware might lead to a revolution for spiking networks, similar to what GPU’s provided for CNN’s