Cognitive Neuroscience and Embodied Intelligence Introduction to Cognitive Neuroscience Based on book Cognition, Brain and Consciousness ed. Bernard J. Baars courses taught by Prof. Randall O'Reilly, University of Colorado, and Prof. Włodzisław Duch, Uniwersytet Mikołaja Kopernika and http://wikipedia.org/ http://grey.colorado.edu/CompCogNeuro/index.php/CECN_CU_Boulder_OReilly http://grey.colorado.edu/CompCogNeuro/index.php/Main_Page Janusz A. Starzyk EE141 1 … from the brain, and from the brain alone, arise our pleasures, joys, laughter and jokes, as well as our sorrows, pains, grief's and tears. Through it, in particular, we think, see, hear, and distinguish the ugly from the beautiful, the bad from the good, the pleasant from the unpleasant… » Attributed to Hippocrates, 5th century BC 2 2 EE141 The Brain ... The most interesting and the most complex object in the known universe How can we understand the workings of the brain? On what level should we attack this question? An external description won’t help much. How can we understand the workings of a TV or computer? Experiments won’t suffice, we must have an understanding of the operating principles. To verify that we understand how it works, we must make a model. 3 EE141 How do we know anything? An important question: how do we know things? Examples: super diet based on dr. K, Chinese medicine and other miracle methods. How do we know that they work? How do we know that they are for real? Gall noticed that the skull shape decides about ones abilities. Thousands of cases confirmed his observations. Craniometry: measuring the bones of the skull to determine intelligence. Do I know or I only believe that I know? Not being certain allows to learn, certainty makes learning difficult. If we know how easy it is to err we could avoid a scientific fallacy. 4 EE141 How to understand the brain? To understand: reduce to simpler mechanisms? Which mechanisms? Analogies with computers? RAM, CPU? Logic? Those are poor analogies. Psychology: first you must describe behavior, it looks for explanations most often on a descriptive level, but how to understand them? Physical reductionism: mechanisms of the brain. Reconstructionism: using mechanisms to reconstruct the brain’s functions To create: what must we know in order to create an artificial brain? We can answer many questions only from an ecological and evolutionary perspective: why is the world the way it is? Because that’s how it made itself ... Why does the cortex have a laminar and columnar structure? 5 EE141 From molecules through neural networks 10-10 m, molecular level: ion channels, synapses, properties of cell membranes, biophysics, neurochemistry, psychopharmacology; 10-6 m, single neurons: neurochemistry, biophysics, LTP, neurophysiology, neuron models, specific activity detectors, emerging. 10-4 m, small networks: synchronization of neuron activity, recurrence, neurodynamics, multistable systems, pattern generators, memory, chaotic behaviors, neural encoding; neurophysiology ... 10-3 m, functional neural groups: cortical columns (104-105), group synchronization, population encoding, microcircuits, Local Field Potentials, large-scale neurodynamics, sequential memory, neuroanatomy and neurophysiology. 6 EE141 … to behavior 10-2 m, mesoscope networks: sensory-motor maps, self-organization, field theory, associative memory, theory of continuous areas, EEG, MEG, PET/fMRI imaging methods ... 10-1 m, transcortical fields, functional brain areas: simplified cortical models, subcortical structures, sensory-motor functions, functional integration, higher psychic functions, working memory, consciousness; (neuro)psychology, psychiatry ... Cognitive effects Principles of interactions Neurobiological mechanisms 7 EE141 … to the mind Now a miracle happens ... 1 m, CNS, the whole brain and organism: An interior world arises, intentional behaviors, goal-oriented actions, thought, language, everything that behavioral psychology examines. Approximations of neural models: Finite State Machine, rules of behavior, models based on the knowledge of cognitive mechanisms in artificial intelligence. What happened to the psyche, the internal perspective? Lost in translation: neurons => networks => behavior 8 EE141 … to the mind “What if … we were magically shrunk and put into someone’s brain while he was thinking. We would see all the pumps, pistons, gears, and levers working away, and we would be able to describe their working completely, in mechanical terms, thereby completely, describing the thought process of the brain. But that description would nowhere contain any mention of thought! It would contain nothing but descriptions of pumps, pistons, levers!’ - Gottfried Leibnitz 1690 9 EE141 Levels of description Summary (Churchland, Sejnowski 1988) Sensing outside and inside the body 10 EE141 Distance – from 10-10 m to one meter Small molecules can change brain functions and resulting behavior. Around year 1800 people were surprised to find out that nitrous oxide (N2O) changes their behavior – it produces small amount of neurotransmitter. 11 11 EE141 Time scales - 10 orders of magnitude Neurons can fire as fast as 1000 Hz. Our brain deals with events on the time range from years to milliseconds. 100ms is about the fastest we can react to an event. – Slower reaction time would prevent humans from protecting themselves from dangers and they would have no chance to survive and reproduce, – faster reaction time would overwhelm the brain to combine sensory inputs and determine the direction and speed of the attacker. Some skills take long time to master like playing guitar or learning how to speak. 12 12 EE141 Making Inferences Inferences explanatory concepts from raw observations - play an important role in science. Figure showing relation between observations: lights seen in the sky, and the inferences drawn, path of the planets around the sun. 13 13 EE141 Working Memory Observations, based on experimental data, are important in cognitive science. Concepts like working memory and their size (7+/-2) are not ‘given’ in nature but are inferred from experimental observations. Emerge from years of testing, working memory proposed after a 2 decade study of immediate memory 14 14 EE141 Working Memory Models EEG (Electroencephalography), fMRI (Functional Magnetic Resonance Imaging), etc are inferential measurements of brain. Results for working memory converge well with behavioral measurement. Combined sources of evidence are widely used for study in cognitive neuroscience 15 15 EE141 Neurocognitive Models Computational cognitive neuroscience: detailed models of cognitive functions and neurons. Neurocognitive computing: simplified models of higher cognitive functions, thinking, problem solving, attention, language, cognitive and behavioral controls. Example models: Self-organization, dynamic net or biophysical spiking neurons. Lots of speculations, but qualitative models explaining the results of psychological experiments as well as the causes of mental illnesses are developing quickly. Even simple brain-like information processing yields results similar to the real ones! Warning against excessive optimism based on behavioral models. 16 EE141 Model of self-organization Topographical representations in numerous areas of the brain: sensory impulses, multimodal maps of orientation, visual system maps and maps of the auditory cortex. o Model (Kohonen 1981): competition between groups of neurons and local cooperation. x=data o=weights of neurons x o o o o x o o x o xo N-dimensional input space o o o Neurons react to signals adjusting their parameters so that similar impulses awaken neighboring neurons. Weights locate points in N-D neural network w 2-D 17 EE141 Dynamic model Strong feedback, neurodynamics. Hopfield model: associative memory, learning based on Hebb’s law, synchronized dynamics, two-state neurons. Vector of input potentials V(0)=Vini , i.e. input = output. Dynamics (iterations) Hopfield’s network reaches stationary states, or the answers (vectors of elemental activation) of the network to the posed question Vini (autoassociation). If the connections are symmetrical then such a network trends to a stationary state (local attractor). Vi t 1 sgn I i t 1 sgn t = discrete time. EE141 j WijV j j 18 Biophysical model – spiking neurons Synapses Soma I syn (t ) Spike EPSP, IPSP Rsyn Csyn Spike Cm Rm “Spiking Neuron Models”, W. Gerstner and W. Kistler Cambridge University Press, 2002 http://icwww.epfl.ch/~gerstner//SPNM/SPNM.html EE141 19 Abstract neuron Add all inputs considering synaptic strength. Neurons activation cannot grow indefinitely: pass the total net input through a sigmoidal limiting function: Output activation does not exceed a unit vale. Is this how neurons respond? No, but this is how the average number of impulses per second changes as a function of neuron’s activation. 20 EE141 Molecular foundations Action potentials are the result of currents which flow through the ionic channels in the cell membrane Hodgkin and Huxley measured these currents and described their dynamics through differential equations. -70mV Na+ Action potential K+ Ca2+ Ions/protein EE141 21 Biological Neural Nets Ions flow through the neurons’ membranes under the forces of electricity and concentration gradients, changing their polarization Vm .Sodium ions Na+, potassium K+, calcium C++, chloride Cl flow to equalize the charge distribution; their imbalance creates the electrical potential which restores the balance. Ions flow through channels finding resistance I = V/R: Conductivity is G=1/R, so I=VG (Ohms law) Dyffusion generates current I in proportion to ion concentration C I = DC (Fick's First law) Equilibrium potential E counteracts diffusion: I = G(VE) EE141 22 Ions and Neurons Glutamic acid opens Na+ channels, (excitatory), GABA works on Cl- channels inhibiting excitation. Liquid outside neurons similar to a sea water and contains: NaCl, KCl. Sodium pomp polarizes membranes: removes sodium idons, and brings in potassium ions. Sodium ions dominate, so the resting potential is + 23 outside and –70mV inside. EE141 Ions and Neurons Glutamic acid opens Na+ channels, (excitatory), GABA works on Cl- channels inhibiting excitation. Liquid outside neurons similar to a sea water and contains: NaCl, KCl. Sodium pomp polarizes membranes: removes sodium idons, and brings in potassium ions. Sodium ions dominate, so the resting potential is + 24 outside and –70mV inside. EE141 Ions and Neurons follows from the sodium pump, which creates the “dynamic tension” for subsequent neural action. Glutamateopens Na+ channelsNa+ enters (excitatory) GABAopens Cl- channelsCl- enters if Vm < threshold (inhibitory) Alcohol closes sodium channels Na. General anesthesia: opens K. Scorpions have various toxins, eg. opens Na, closes K. Everything 25 EE141 Putting it Together To find potential Vm for each ion one needs to consider its equilibrium potential Ec, gc(t) is a fraction of ion channels that are open in a given moment, ĝc is max. conductivity of all channels; therefore a produce ĝc gc(t) gives us conductivity. Considering diffusion, current for a given ion is: Ic = ĝc gc(t) (Vm(t)Ec) For equilibrium potential Ic=0 Total current for 3 most important channels: 26 EE141 It's Just a Leaky Bucket Good analogy: ge rate of flow into bucket; gi and gl rate of “leak” out of bucket. Resulting water level? Vm balance between these forces. Or like a tug-of-war (rope pooling) 27 EE141 Two inputs As a result of the current flow Inet potential changes with some time delay dtvm: Two excitatory inputs at time t =10, assuming conductances ge=ĝege(t) = 0.2 i 0.4 and gl=2.8 Current flows, buts stops at equilibrium. If there are constant excitations (open channels) neuron reaches new equilibrium potential. 28 EE141 Overall Equilibrium Potential If Inet=0 then we can computeVm (normalized threshold 0.25): Can now solve for the equilibrium potential as a function of inputs. Simplify: ignore inhibition for a moment, set Ee=1 a Ei=0 (leaks always on El=1) Membrane potential computes a balance (weighted average) of excitatory and inhibitory inputs. 29 EE141 Computational Neurons (Units) Summary Use bias b and activation threshold Q to compute the output signal with the total number of input connections N ([ . ]+ is a positive part): Weights = synaptic efficacy; weighted input = xiwij. Net conductances (average across all inputs) excitatory (net = ge(t)), inhibitory gi(t). Function gx/(gx+1) combined with Gassian noise is similar to sigmoidal, Parameter g regulates the slope. 30 EE141 Thresholded Spike Outputs Neuron’s behavior is a result of currents equilibrium in various channels. 31 EE141 Parameters Emergent allows to simulate neurons with realistic parameters. 32 EE141 Emergent: Emergent is a powerful tool for simulation of biologically plausible, complex neural networks: http://grey.colorado.edu/emergent Emergent supports: Simulating the brain functions Classic back-propagation and recurrent back-propagation and variants, Constraint-satisfaction (CS) including the Boltzmann Machine, Interactive Activation and Competition, and other related algorithms; Self-organized learning including Hebbian Competitive learning and variants, with Kohonen's Self-Organizing Maps and variants Leabra (``local error-driven and biologically realistic algorithm'') Real Time Neural Simulator Long Short Term Memory Oscillating Inhibition Learning Mechanism 33 EE141 Emergent and Other Simulators: Emergent started as PDP (parallel distributed processing) developed by McClelland and Rumelhart in 1986 Other simulators: GENESIS NEURON NEST XPPAUT SPLIT Mvaspike SNNS Topographica NMS FANN Commercial simulators: Matlab NN toolbox Mathematica NN package Peltarion Synapse Robotics simulations with rigid body physics Emergent is compared to other neural network simulators at: http://grey.colorado.edu/emergent/index.php/Comparison_of_Neural_Network_Simulators 34 EE141 Emergent: Unit.Proj Emergent allows to simulate neurons with realistic parameters. act_fun activation function: Noisy xx1; can be chosen without noise, linear with or without noise, or spike. g_bar_e determines fraction of channels is open during on/off_cycle. e_rev_e is equilibrium potential for activating channels. a = accommodation, increase of Ca++ concentration in neuron => opens inhibitory channels (usually K+) – section 2.9. h = hysteresis, reaction slowdown, active neurons remain active for some time after the excitation is removed. 35 EE141 Emergent: results Net = total excitation changes from 0 to g_bar_e=1 (all channels open). I_net: current flows to the neuron, equilibrium is reached, then it flows out of the neuron. V_m is axon potential, increases from -70mV (here 0.15) to +50mV (here 0.30). Act = output activation; if spikes are selected then they will show on the figure Single impulses; fluctuations result form noise, here there is a small noise variance = 0.001 Act eq = equivalent of average rate-code. EE141 36 Advantages of model simulations Models help to understand phenomena: enable new inspirations, perspectives on a problem allow to simulate effects of damages and disorders (drugs, poisoning) help to understand behavior models can be formulated on various levels of complexity models of phenomena overlapping in a continuous fashion (e.g. motion or perception) models allow detailed control of experimental conditions and exact analysis of the results Models require exact specification of underlying assumptions: allow for new predictions perform deconstructions of psychological concepts (working memory?) allow to understand the complexity of a problem allow for simplifications enabling analysis of a complex system provide a uniform, cohesive plan of action 37 EE141 Disadvantages of simulations One must consider limitations of designed models: Models are often too simple, they should contain many levels. Models can be too complex, theory may give simpler explanation why there are no hurricanes on the equator? - due to Coriolis effect It’s not always known what to provide for in a model. Even if models work, that doesn’t mean that we understand the mechanisms. Many alternative yet very different models can explain the same phenomenon. Models need to be carefully designed to fit the observations: What’s important in building a model are general rules the more phenomena a model explains, the more plausible and universal it is. Allowing for interaction and emergence (construction) is very important. Knowledge acquired from models should undergo accumulation. 38 EE141 Cognitive motivation Although the thinking process seems to be sequential information processing, more detailed models predict parallel processing Gradual transition between conscious and subconscious processes Parallel processing of sensory-motor signals by tens of millions of neurons Specialized areas of memory responsible for various representations e.g. shape, color, space, time Levels of symbolic representation More diffuse than binary logic Learning mechanisms as a foundation for cognitive science When you learn, you change the method of information processing in your brain Resonance between ”bottom-up” representation and ”top-down” understanding Prediction and competition of ideas 39 EE141 Brain Landmarks Most terms in neuroscience are Latin names, as it was the language of science. Medial (midline) view of the brain also called mid-sagittal section of the brain is a slice from the nose to the back of the head. Corpus callosum is a fiber bridge flowing between right and left hemispheres, begins behind the frontal lobe and loops up and ends in front of the cerebellum. 40 40 EE141 Brain Landmarks Lateral (side) view of left hemisphere is shown here. Folds in the cortex are important part of anatomy. Longitudinal fissure runs along the midline between right and left hemispheres. Lateral sulcus runs forward at a slant along the side of the brain and divides the temporal lobe from the main cortex. Central sulcus divides the rear half (posterior half) of the brain from the frontal lobe. EE141 41 41 Brain Landmarks Temporal lobe points in the direction of the eye. The three major planes of section (cuts) are: Vertical section (sagittal) from the front of the brain to the back. – Slice through the midline is called midsagittal. Horizontal slice. Coronal section (named for its crown shape). 42 EE141 Body Landmarks The three major planes of section : Vertical section (sagittal) from the front to the back. Horizontal (transverse) section. Frontal (coronal in the brain) section. Other important directions: Superior (dorsal) and inferior (ventral) Medial and lateral Anterior (rostral) and posterior (caudal). EE141 43 Mind and Brain Visual perception: viewing natural imagery we must understand ways of encoding objects and scenes. Spatial awareness: considering the interaction between streams of visual information will let us simulate concentration Memory: modeling hippocampal structures allows us to understand various aspects of episodic memory, and learning mechanisms show how semantic memory arises. Working memory: explaining the capacity to simultaneously hold in the mind several numbers, while performing calculations requires specific mechanisms in the neural model. 44 EE141 Mind and Brain Reading words: the network model in Emergent will learn to read and pronounce words and then to generalize its knowledge to the pronunciation of new words as well as to recreate certain forms of dyslexia. Semantic representations: analyzing a text on the basis of context, the appearance of individual words, the network will learn the semantics of many ideas. Decision-making and task execution: A model of the prefrontal cortex will be able to keep attention on performed tasks in spite of hindering variables. Development of the representation of the motor and somatosensory cortex: through learning and controlled selforganization; 45 EE141 Mind and Brain Andreas Vesalius (1514-1564), a Belgian physician, published the first known detailed anatomy based on dissections of human body. He showed that both men and women have the same number of ribs. Illustrations, like the brain shown here, were done by Titian. 46 46 EE141 Mind and Brain Paintings, like the Rembrandt (The Anatomy Lesson of Dr.Tulp), show the excitement generated by dissection of human cadavers. René Descartes (1596 -1650) a mathematician and philosopher is considered as the originator of modern mind/body philosophy. He said most famously, cogito ergo sum ("I think, therefore I am"). Thinking is thus every activity of a person of which he is immediately conscious. Descartes' "error" pointed by António R. Damásio was the separation of mind and body. 47 47 EE141 Mind and Brain Charles Darwin (1809 –1882) wrote a book “Expression of emotions in man and animals” pointing towards biological origins of emotions and not just cultural as people thought. He also stressed the importance of culture and environment, that helps to resolve “nature vs nurture” debate. 48 48 EE141 Mind and Brain Santiago Ramon y Cajal (1852–1934) founder of brain science studied properties of neurons. He observed neurons under microscope and showed that they are single cells that end with synapses Nerve impulses travel down the axon to synapses In 1952 Hodgkin and Huxley constructed action potential model for a spiking neuron 49 49 EE141 Neurons Cajal’s drawing of a slice of chicken brain exposed using Gogli staining method A Modern version 50 EE141 Mind and Brain Pierre Paul Broca (18241880) discovered the region in the brain responsible for speech production In 1861 he studied a patient with epilepsy who lost ability to speak On the patient’s death Broca performed autopsy and showed a damage to the posterior part of the third frontal convolution in the left hemisphere and associated it to speech production Much of what we know about brain was first discovered by studying various deficits 51 51 EE141 Mind and Brain Wernicke’s area (W), in the left upper part of the temporal lobe, is an important area for receptive language (understanding). Carl Wernicke (1848-1905) published his finding shortly after Broca’s work The two areas are connected for speech comprehension and production. Damage (in or near) leads to: Broca’s area (B): Expressive aphasia, Wernicke’s area (W): Receptive aphasia, Fibers between B & W: Disconnection aphasia. 52 EE141 Mind and Brain Left hemisphere is responsible for language production and listening while right hemisphere is concerned with emotional aspects of language. Angelo Mosso (1846-1910), found a way to measure blood pressure during demanding mental tasks. Mosso’s work anticipated current measures of brain blood flow like fMRI. fMRI measures local blood flow changes in the brain. The fMRI responds to blood flow changes whenever some brain regions require more oxygen and glucose. EE141 53 53 Mind and Brain Nineteen century scientists were very interested in consciousness. William James (1890) declared psychology as a science of conscious mental life. Many scientists (Helmholtz, Loeb, Pavlov) disagreed – they took on a physicalistic view of mental life. Pavlov experiments with dogs (1900) on classical conditioning convinced psychologists that all behavior can be derived from simple reflexes. 54 54 EE141 Mind and Brain In 1970-ies many scientists were dissatisfied with behaviorism. Different methods of testing conscious and unconscious brain events were developed Figure compares results of study using visual backward masking method based on fMRI to compare brain activity for conscious and unconscious visual words. 55 55 EE141 Conclusion Ongoing debates in cognitive neuroscience: Local vs distributed functions in the brain The question of consciousness Unconscious inferences in vision Capacity limits in the brain Short-term and long-term memory – separate or not? The biological basis of emotions Nature vs nurture – genes vs environment Cognitive neuroscience combines psychology, neuroscience and biology to answer questions about mind and brain. Modeling cognitive functions of the brain helps to understand psychological phenomena and predict behavior. It may simplify complex cognitive processing with full control of experimental conditions. It helps to build working models of embodied intelligence 56 56 EE141